electromechanical energy conversion in asymmetric...
TRANSCRIPT
Electromechanical Energy Conversion
in Asymmetric Piezoelectric
Bending Actuators
Vom Fachbereich Mechanik
der Technischen Universitat Darmstadt
zur Erlangung des Grades eines
Doktor-Ingenieurs (Dr.-Ing.)
genehmigte
Dissertation
von
Dipl.-Ing. Kai-Dietrich Wolf
aus Wiesbaden
Referent: Prof. Dr. P. Hagedorn
Korreferent: Prof. Dipl.-Ing. Dr. techn. H. Irschik
Tag der Einreichung: 26.04.2000
Tag der mundlichen Prufung: 29.05.2000
Darmstadt 2000
D 17
Vorwort
Die vorliegende Arbeit entstand wahrend meiner Tatigkeit als wissenschaftlicher
Mitarbeiter bei Prof. Dr. P. Hagedorn in der Arbeitsgruppe Dynamik des Fach-
bereichs Mechanik der Technischen Universitat Darmstadt.
An erster Stelle mochte ich mich bei Herrn Professor Hagedorn bedanken fur
die Anregung zu dieser Arbeit und die durch seine Personlichkeit gesetzten
akademischen und menschlichen Rahmenbedingungen, die das Arbeiten in dieser
Gruppe uberaus angenehm machten. Herrn Professor Irschik gilt mein Dank fur
die bereitwillige und sehr wohlwollende Ubernahme des Korreferats.
Besonders herzlich bedanke ich mich bei meinen ehemaligen Kolleginnen und
Kollegen, Frau Jutta Braun und Frau Renate Schreiber, den Herren Dipl.-Ing.
Thomas Sattel, Marcus Berg, Daniel Sauter, Georg Wegener, Hartmut Bach,
Norbert Skricka, Roland Platz, Malte Seidler, Uli Ehehalt, den Herren Dr.-Ing.
Minh Nam Nguyen, Joachim Schmidt, Ulrich Gutzer, Thomas Hadulla, Karl-
Josef Hoffmann, Dirk Laier, Utz von Wagner, Christoph Reuter und Herrn
Prof. Dr.-Ing. Wolfgang Seemann fur Ihre Hilfsbereitschaft und stetige Anteil-
nahme.
Ein wesentlicher Teil dieser Arbeit wurde von meinen Diplomarbeitern, den Her-
ren Dipl.-Ing. Stephan Frese und Boris Stober mitgeleistet. Ihnen danke ich fur
ihren Fleiß und die sehr freundschaftliche Zusammenarbeit.
Die Herren Professoren A.F. Ulitko und Oleg Yu. Zharii haben durch ihre Anre-
gungen dieser Arbeit wesentliche Impulse gegeben. Auch fur die mir entgegenge-
brachte menschliche Nahe mochte ich ihnen an dieser Stelle besonders danken.
Darmstadt, im Juni 2000
Kai Wolf
Contents
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Scope of this Work . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Piezoelectric Ceramics 3
2.1 A Model for Piezoelectric Solids . . . . . . . . . . . . . . . . . . . 3
2.1.1 A derivation of HAMILTON’s principle for dielectric solids . 4
2.1.2 Constitutive equations . . . . . . . . . . . . . . . . . . . . 8
2.1.3 Electrical boundary conditions . . . . . . . . . . . . . . . 10
2.1.4 Engineering notation . . . . . . . . . . . . . . . . . . . . . 10
2.2 Piezoceramic Solids . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 The Electromechanical Coupling Factor . . . . . . . . . . . . . . 12
2.3.1 IEEE/IRE standards definitions . . . . . . . . . . . . . . 16
2.3.2 Dynamic/Effective EMCF . . . . . . . . . . . . . . . . . 18
II Contents
3 Optimal Design 21
3.1 The Laminate Beam . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 Maximum EMCF . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Maximum Dynamic EMCF . . . . . . . . . . . . . . . . . . . . . 27
3.4 Maximum Power Factor . . . . . . . . . . . . . . . . . . . . . . . 30
4 Verification 35
4.1 EMCF: Theory versus Finite Element Model for a periodic system 35
4.2 Power Factor: Theory versus Experiment for an infinite beam . . 39
4.2.1 The problem geometry . . . . . . . . . . . . . . . . . . . . 39
4.2.2 Model for the bonded actuator . . . . . . . . . . . . . . . 39
4.2.3 Experimental setup . . . . . . . . . . . . . . . . . . . . . . 47
4.2.4 Optimal patch geometry . . . . . . . . . . . . . . . . . . . 51
5 Discussion and Conclusion 55
5.1 Discussion of the results . . . . . . . . . . . . . . . . . . . . . . . 55
5.2 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Chapter 1
Introduction
1.1 Motivation
Among the various wave types in solid structures, bending waves are associated
with the largest deflections. Consequently, most of the mechanical interaction
with the surrounding environment, e.g. sound radiation, is to be accounted to
the bending or flexural deformation [CH88], typically observed in plate or beam
structures which are characterized by a small thickness compared to longitudi-
nal dimensions. The large attainable deflections and the resulting potential for
mechanical interaction is the motivation to control flexural waves in beams and
plates. Control strategies are often implemented in electronic circuits or micro-
processors and the power required to control the structure will most likely be
provided electrically. Generally, actuators are the means required to convert the
electrical power supplied into any form of mechanical power. Ideally, the me-
chanical actuation would be suitable to enforce control authority over all flexural
waves in the structure. In structures containing beam or plate members, this
requires several distributed low authority actuators rather than a single one with
high control authority and structures with integrated distributed actuators are
often referred to as intelligent or smart structures [CdL87]. Patches of piezo-
electric ceramics, which can easily be bonded to the surface of beams or plates,
have proven to be well suited to serve as actuators to control flexural vibrations
and optimization of the design [KJ91] of these patches for this purpose has been
a concern.
2 Chapter 1. Introduction
In high power applications, such as ultrasonic traveling wave motors, where
piezoceramic layers are employed for continuous energy transfer [UTKN93], the
power flow and thus the maximum power output of the motor is limited by
saturation effects. Actuators should be designed in a way that critical field
variables are kept as small as possible. Power loss, which can be caused by a
variety of loss mechanisms in the actuator, leads to undesired heat production.
Material properties of the piezoelectric material are temperature sensitive and
excessive heat can even cause permanent depolarization. Therefore, the design
of the piezoelectric bending actuator should be carefully considered to provide
optimal energy conversion.
1.2 Scope of this Work
Once the need for optimization is recognized, the first question to ask is: what is
optimal? The more details about the particular structure are known, the more
comprehensive but also complex can the mathematical model be and analytical
solutions will generally not be available. In the case of the ultrasonic traveling
wave motor, energy is transferred between rotor and stator by a highly nonlinear
friction mechanism which defies detailed analytical treatment. An optimization
criterion would ideally be independent of such detailed modeling and still give
authoritative guidelines for the design of the actuator.
This work aims at a general consideration of energy conversion in bending
actuation rather than detailed modeling of particular systems. Energy conver-
sion will be considered good if a large fraction of the electrical work supplied to
the system is converted into mechanical work and vice versa, where it is acknowl-
edged that energy conversion has to be carried out in cycles. It will be shown that
the electromechanical coupling factor (EMCF) and the actuator power factor are
measures which comply with this definition, based on a hypothetical quasi-static
work cycle or harmonic steady state vibration, respectively. In spite of the fact
that power actuators are most likely operated in the regime of nonlinear material
behavior, the analysis carried out is confined to the domain of linear equations.
This restriction is due to the difficulties arising in the analytical treatment of
nonlinear material behavior in continuous systems. It is anticipated here, that
an optimization of the energy conversion based on linear assumptions will lead
to an overall reduction of field intensities in the piezoelectric material and thus
increase the attainable power flow and reduce the power loss due to various loss
mechanisms.
Chapter 2
Piezoelectric Ceramics
2.1 A Model for Piezoelectric Solids
The term ’dielectric solid’ in this work is used for a continuous elastic body
with dielectric properties, which constitute the interaction between the electri-
cal quantities dielectric displacement and electric field. As in the case of the
elastic interaction between strain and stress it is assumed that the dielectric
interaction is free of loss mechanisms. A dielectric solid which also exhibits
lossless interaction between electrical and mechanical quantities will be referred
to as piezoelectric solid, so piezoelectric solids represent a subset of the dielec-
tric solids. The absence of loss implies that the state of the system is uniquely
determined by the independent state variables, regardless of the path taken in
state space. Feasible paths can mathematically be determined as solutions to
the equations of motion.
An elegant method to obtain the equations of motion is the use of varia-
tional principles, which also have certain advantages when complicated bound-
ary conditions are to be treated. For piezoelectric solids, Hamilton’s principle
is frequently employed and instead of the internal energy U which is used for me-
chanical problems, the electric enthalpy H is substituted into the Lagrangeian.
Many authors refer to Tiersten [Tie69], who probably coined the name ’electric
enthalpy’, to explain the origin of this expression, but no satisfactory derivation
can be found in this reference. Variational methods for piezoelectric solids were
presented by Tiersten, Holland and Eer Nisse earlier [Tie68], [HN68],
4 Chapter 2. Piezoelectric Ceramics
[Nis67], but there it is rather demonstrated that the requirement of a particular
energy expression to be stationary leads to the balance equations, than how to
arrive at the electric enthalpy H and a corresponding expression for the virtual
work for substitution into Hamilton’s principle. Therefore, a brief derivation
of Hamilton’s principle shall be given and it will be seen that the electric
enthalpy density H but also the energy density U can be substituted into the
Lagrangeian, provided a corresponding expression for the virtual work is used.
2.1.1 A derivation of HAMILTON’s principle for dielectric solids
The balance of momentum for a continuum can be written as
Tij,j + Fi =d
dt(ρui), (2.1)
where Tij , Fi and ui are the components of the stress tensor, the force per unit
volume and the velocity, respectively, and ρ is mass per unit volume. For the
applications considered in this work, interaction between electrical quantities
can be treated as quasistatic [Mau88], [Aul81], so magnetic field effects can be
neglected and Maxwell’s equations reduce to
Dk,k = ρf, (2.2)
Ek = −φ,k, (2.3)
for the electric displacement Dk and the electric field Ek, which can be repre-
sented by the negative gradient of the electric potential φ. Here, the comma
indicates differentiation with respect to the orthogonal cartesian coordinates
xi , (i = 1, 2, 3). Inside the electrically insulating dielectric, free charges do not
exist and the free charge density per unit volume ρf in (2.2) is zero. However,
free charges will exist on the electroded surfaces of the dielectric.
The above equations of equilibrium (2.1)-(2.3) hold at any point of a conti-
nuum1) and from now on, a volume fraction of this continuum which is occupied
by the dielectric solid, having volume V , is considered. On the surface S of the
solid, mechanical and electrical boundary conditions apply. Mechanical bound-
ary conditions will be given on the distinct sections Su, Sf of the surface either
in terms of prescribed displacement
ui = ui on Su, (2.4)
1)Provided, that the quasistatic approximation is justified, of course.
2.1. A Model for Piezoelectric Solids 5
or prescribed force per unit area
Tijnj = fi on Sf , (2.5)
where nj is the outward normal unit vector and Su ∪ Sf = S. Correspondingly,
electrical boundary conditions are either prescribed surface charge density
−Dini = σ on Sσ, (2.6)
or prescribed potential
φ = φ on Sφ, (2.7)
and Sφ ∪ Sσ = S. The free charges on the surface electrodes can be represented
by a density σ of charge per unit surface, if the thickness of the electrode is
negligible. This, and the assumption that the dielectric displacement on the
outside of the dielectric body is small compared to the inside in connection with
(2.2), leads to (2.6).
In a similar way as it can be done for a purely mechanical system [Was68],
a variational principle can be derived from the equations of equilibrium (2.1)-
(2.3) and boundary conditions (2.4)-(2.7). After multiplication by the variation
δui, δφ of the quantities ui, φ and integration over the volume and surface of the
body, a weak form
−∫V
[Tij,j + Fi − d
dt(ρui)
]δui dV +
∫Sf
(Tijnj − fi)δui dS
−∫V
Dk,kδφdV +
∫Sσ
(Dini + σ)δφdS = 0 (2.8)
of the problem (2.1)-(2.7) is obtained. Here, the choice of the signs of the integral
expressions in (2.8) is arbitrary, but in this form particularly suitable for the
further derivation. The variations δui, δφ are chosen to satisfy the displacement
and potential boundary conditions, i.e. they vanish on Su and Sφ, respectively.
Partial integration and application of the divergence theorem lead to∫V
(TijδSij −DkδEk) dV −∫V
[Fi − d
dt(ρui)
]δui dV
+
∫Sσ
σδφdS −∫Sf
fiδui dS = 0, (2.9)
where the linear strain tensor
Sij =1
2(ui,j + uj,i) (2.10)
6 Chapter 2. Piezoelectric Ceramics
has been introduced, assuming that only small deformations are encountered.
If only contributions of the mechanical and electrical quantities to the energy
of the system are considered, the total differential
dU = Tij dSij + Ek dDk (2.11)
of the internal energy density U , shows that U is a function of the indepen-
dent variables strain Sij and dielectric displacement Dk. Let another energy
expression, the electric enthalpy density H [Tie69], be defined as
H = U − EkDk, (2.12)
which is obtained from U by a Legendre transformation. Since
dH = Tij dSij −Dk dEk, (2.13)
H is a function of the independent variables strain Sij and electric field Ek which
in turn are functions of ui and φ, respectively. If the volume forces Fi vanish
and ρ is independent of time, (2.9) can be written∫V
δH dV +
∫V
ρd
dt(ui)δui dV +
∫S
(σδφ− fiδui) dS = 0, (2.14)
and further partial integration over time between t1 and t2 leads to∫ t2
t1
∫V
[δH − 1
2ρ δ(u2i )
]dV dt+
∫ t2
t1
∫S
(σδφ− fiδui) dS dt = 0, (2.15)
requiring the variations δui and δφ to vanish at t1, t2. With the Lagrangeian
L =
∫V
(1
2ρu2i −H) dV =
∫V
(T −H) dV, (2.16)
where T denotes kinetic energy density of the body, and the virtual work of the
given forces fi and charges σ
δW =
∫S
(fiδui − σδφ) dS, (2.17)
we obtain Hamilton’s principle for the electromechanic continuum
δ
∫ t2
t1
L(ui, φ) dt+
∫ t2
t1
δW (ui, φ) dt = 0, (2.18)
taking into account electrical and mechanical quantities. This principle states,
that changes of theses quantities in a continuum are possible only in a manner
2.1. A Model for Piezoelectric Solids 7
that satisfies (2.18). If it is now assumed, that displacement ui and electric po-
tential φ are independent state variables of a dielectric solid which are sufficient
to determine the state of the solid uniquely2), then the possible states of the
dielectric solid have to satisfy (2.18). There is a unique mapping of the state of
the dielectric solid onto the electric enthalpy density H, and this function will
determine the behavior of the dielectric solid, so (2.18) can be referred to as
Hamilton’s principle for dielectric solids.
In some applications it might be preferable to choose displacement ui and
surface charge density σ = −Dini as independent quantities. As shown above,
dependent and independent state variables in the energy expressions L and W
can be interchanged by a Legendre transformation, leading to a complementary
form of Hamilton’s principle
δ
∫ t2
t1
L(ui,Di) dt+
∫ t2
t1
δW (ui,Di) dt = 0 (2.19)
where the transformed Lagrangeian
L =
∫V
(T −H − EiDi) dV =
∫V
(T − U) dV (2.20)
and a corresponding expression for the virtual work of external forces fi and
electric potential φ
δW = δW +
∫S
δ(σφ) dS =
∫S
(fiδui + φδσ) dS (2.21)
have been introduced. The expressions (2.18) and (2.19) are equivalent, since
δ
∫V
EkDk dV = −δ∫V
φ,kDk dV = −δ∫S
φDknk dS + δ
∫V
φDk,k dV
=
∫S
δ(φσ) dS, (2.22)
taking into account, that the density of free charges in the dielectric solid is
zero and therefore Dk,k = 0. It should be mentioned here, that the expression
(2.19) is not a weak form of the problem (2.1)-(2.7), i.e. if the mathematical
transformations applied in the course of the derivation of Hamilton’s principle
in the form (2.18) are employed in reverse sequence, (2.19) does not lead to
the equilibrium equations (2.1), (2.2) [Was68]. Still, a variational principle in
the form (2.19) can be useful to find approximate solutions, e.g. employing the
Rayleigh-Ritz or Galerkin method.
2)This assumption will generally be based on observation [Rei96].
8 Chapter 2. Piezoelectric Ceramics
2.1.2 Constitutive equations
Piezoelectric solids exhibit a strong coupling between electrical and mechanical
state variables. These state variables are interrelated via constitutive equations,
which can be written in differential form
dTij =∂Tij∂Skl
dSkl +∂Tij∂Ek
dEk,
dDi =∂Di
∂SjkdSjk +
∂Di
∂EjdEj , (2.23)
for strain and electric field as independent quantities, if all state changes are
reversible [Rei96]. This implies the material is lossless, i.e. elastic and free
of dielectric or piezoelectric losses. As a consequence of (2.23), there exists a
potential U for the material, which can be represented in differential form by
(2.11). The total differential of the electric enthalpy density H in the general
form
dH =∂H
∂SijdSij +
∂H
∂EidEi, (2.24)
compared with (2.13) leads to
Tij =∂H
∂Sij, Di = − ∂H
∂Ei, (2.25)
and the equations (2.23) become
dTij =∂2H
∂Sij∂SkldSkl +
∂2H
∂Sij∂EkdEk,
dDi = − ∂2H
∂Ei∂SjkdSjk − ∂2H
∂Ei∂EjdEj . (2.26)
For the case of linear material behavior, the partial derivatives in (2.26)
reduce to constants and the constitutive equations
Tij = cEijklSkl − ekijEk,
Di = eiklSkl + εSikEk, (2.27)
are obtained for Ei and Sij as independent variables. The superscripts E and S
indicate that the corresponding constants have to be determined under constant
electric field or strain conditions, respectively. Linear constitutive equations for
2.1. A Model for Piezoelectric Solids 9
piezoelectric crystals were first given by Voigt [Voi28]. From (2.26) and (2.27)
it is obvious, that in the linear case the electric enthalpy density reads
H =1
2Sijc
EijklSkl − EieijkSjk − 1
2Eiε
SijEj (2.28)
and the internal energy density
U = H +EiDi =1
2Sijc
EijklSkl +
1
2Eiε
SijEj (2.29)
is free of any terms involving coupling of electrical and mechanical quantities. If
U is expressed in terms of Ei and Tij
U =1
2Tijs
EijklTkl +EidijkTjk +
1
2Eiε
TijEj , (2.30)
coupling terms do occur. As in (2.26), linear constitutive equations can be
obtained from the above expression in the form
Sij = sEijklTkl + dkijEk,
Di = diklSkl + εTikEk. (2.31)
In turn, for this set of constitutive equations, the corresponding expression for
the electric enthalpy density
H = U − EiDi =1
2Tijs
EijklTkl −
1
2Eiε
TijEj (2.32)
is free of coupling terms. The expressions (2.29) and (2.32) could mislead to
the conclusion that upon their substitution in this form into the respective vari-
ational principles (2.18) and (2.19), the interaction between electrical and me-
chanical quantities would disappear. Whereas the change of the internal energy
density δU and the electric enthalpy δH can apparently be derived from the
corresponding potential functions U and H, this is generally not the case for
the expressions δW and δW . These represent the virtual work, which is per-
formed on the dielectric solid under a virtual change of the independent state
variables. The prescribed force fi, potential φ and charge σ will generally be
non-conservative i.e. arbitrary functions of time and they will not be related to
the respective variations δui, δσ and δφ by a potential function. Consequently,
in the variational principles (2.19) and (2.18) the variations have to be carried
out in the quantities given in the expressions for the virtual work and (2.29),
(2.32) are then not suitable for substitution into these principles.
10 Chapter 2. Piezoelectric Ceramics
2.1.3 Electrical boundary conditions
In almost all of the technical applications of piezoelectric material, electrical
boundary conditions will be prescribed on distinct electroded surfaces. On each
of these surfaces, the potential φ will be constant over the area occupied by the
electrode. Let the surface of the piezoelectric body be covered by N electrodes
Sm , (m = 1...N), of which k are subject to prescribed potential and the potential
on these electrodes be denoted by φm, (m = 1...k) for the prescribed case, and
φm, (m = k+ 1...N) for the free electrodes. On the k electrodes with prescribed
potential φm, no variation δφ is admitted. For the remaining N − k electrodes
with free potential, the variation δφ in the surface integral∫Sσ
(Dini + σ)δφdS =
N∑m=k+1
∫Sm
(Dini + σ)δφm dS (2.33)
of expression (2.8) is not unconstrained, but has to be constant over the surface
Sm, such that
N∑m=k+1
∫Sm
(Dini + σ)δφm dS =
N∑m=k+1
δφm
[∫Sm
Dini dS +
∫Sm
σ dS
](2.34)
and if we write for the total charge Qm of each free electrode Sm∫Sm
σ dS = Qm; (m = k + 1...N), (2.35)
the boundary condition (2.6) is replaced by an integral form∫Sm
Dini dS +Qm = 0; (m = k + 1...N), (2.36)
which determines the unknown potential φm on the N − k free electrodes. On
the unelectroded part of the surface, the surface charge density is assumed to be
equal to zero. Consequently, the partition Sσ of the total surface S of the piezo-
electric body consists of free electrodes and the unelectroded part, whereas the
remaining k electrodes represent Sφ. Internal electrodes will not be considered
here.
2.1.4 Engineering notation
In the majority of the publications treating the computation of piezoelectric
problems, a compressed matrix notation, often referred to as ’engineering nota-
tion’ is used. Making use of symmetry properties of the elastic and piezoelectric
2.2. Piezoceramic Solids 11
tensors, the number of indices in the constitutive equations is reduced according
to the following scheme [IEE88]:
ij or kl p or q
11 → 1
22 → 2
33 → 3
23 or 32 → 4
31 or 13 → 5
12 or 21 → 6
Now, the constitutive equations (2.27) can be written in matrix form
Tp = cEpqSq − ekpEk,
Di = eiqSq + εSikEk, (2.37)
where for the stress tensor
Tij → Tp, (2.38)
the coefficients simply are rearranged in a one dimensional array. For the strain
tensor, the transformation
Sij → Sp, for i = j, p = 1, 2, 3
2Sij → Sp, for i = j, p = 4, 5, 6 (2.39)
is employed, leading to analogous expressions for the internal energy density
dU = Tij dSij + Ek dDk → dU = Tp dSp + Ek dDk (2.40)
in the two systems [HN69]. Depending on the choice of the set of constitutive
equations, some of the respective piezoelectric and elastic matrices have to be
adapted accordingly, while others simply undergo the index reduction indicated
above [IEE88].
2.2 Piezoceramic Solids
The interaction between mechanical and electrical field quantities in piezoelectric
crystals is due to the lack of a center of symmetry in the charge distribution of
12 Chapter 2. Piezoelectric Ceramics
the crystal unit cell. Therefore, piezoelectric crystals are inherently anisotropic
in their material properties [JS93]. What we consider a piezoceramic solid is ini-
tially a composition of a large number of randomly oriented piezoelectric crystals
and isotropic on the macroscopic scale. Exposition to strong electric fields leads
to partial reorientation of the crystal axes and the material is then polarized
into a preferred direction [JCJ71]. As a consequence, polarized piezoceramic
material exhibits planar isotropic behavior and the number of independent pa-
rameters, that constitute the linear material properties on the macroscopic scale,
is significantly reduced compared to the most general case for piezoelectric crys-
tals. Commonly, the axis of polarization is assumed to be the z- or 3-axis. The
choice of the set of independent variables depends on the electrical and mechan-
ical boundary conditions. For the models considered in this work, only one of
the stress components is non-vanishing and constitutive equations of the form
S1 = sE11T1 + sE12T2 + sE13T3 + d31E3,
S2 = sE12T1 + sE11T2 + sE13T3 + d31E3,
S3 = sE13T1 + sE13T2 + sE33T3 + d33E3,
S4 = sE44T4 + d15E2,
S5 = sE44T5 + d15E1,
S6 = 2(sE11 − sE12)T6,
D1 = εT11E1 + d15T5,
D2 = εT11E2 + d15T4,
D3 = εT33E3 + d31(T1 + T2) + d33T3, (2.41)
are convenient to use. Of course, there are three other possible sets of linear
constitutive equations for piezoceramic solids with (Ti,Dj), (Si,Dj) or (Si, Ej)
as independent quantities [BCJ].
2.3 The Electromechanical Coupling Factor
The Electromechanical Coupling Factor k of a piezoelectric crystal was first
introduced by Mason as ”the square root of the ratio of the energy stored in
mechanical form, for a given type of displacement, to the total input electrical
energy obtained from the input battery”. Despite of the generality of this def-
inition, the expression for the Electromechanical Coupling Factor (EMCF) in
[Mas50] is merely a material coupling factor, which can be obtained from the
2.3. The Electromechanical Coupling Factor 13
above definition under the assumption of homogeneous deformation of the piezo-
electric solid 3). The analysis of equivalent circuits gave rise to the establishment
of a relation between resonance and antiresonance frequencies of free vibrating
piezoelectric transducers and the EMCF. This relation is often referred to as the
dynamic [BCJ] or effective [Hun54] EMCF for vibrating transducers and will
be discussed in section 2.3.2. Widely used is the definition for the EMCF given
by the IEEE [IEE88], based on a quasistatic deformation cycle rather than en-
ergy expressions for a particular electroelastic state as it was the case in earlier
definitions by the IRE [IRE58]. These definitions will be considered in section
2.3.1. To clarify some of the terminology used, in particular the terms effective
and material coupling factor, the definition of the EMCF used in this work is
introduced first.
The amount of energy which is converted by a thermodynamic system from
one form into another depends on the path, which the state variables take in
state space. So naturally, information about just one particular state cannot
be sufficient to judge a system’s ability to transform energy. To compare this
ability for different paths and systems, a very intuitive restriction is to require
the paths to be closed loops, so the energy conversion process can be carried out
repeatedly and the system reaches its initial state after each cycle. A very well
known closed-loop energy conversion process in thermodynamics is the Carnot
cycle, which was developed to investigate the efficiency of machines operating
in cycles [Rei96]. For piezoelectric transducers, feasible cycles in state space
will be determined by Hamilton’s principle (2.18) and a corresponding electric
enthalpy density, e.g. (2.28).
The EMCF used in this work is based on a quasi-static cycle of deformation
and was introduced by Ulitko [Uli77]. It is defined according to
k2 =Uconv
Uoc=Uoc − Usc
Uoc, (2.42)
as the square root of the ratio of the convertible to the total internal energy of
the structure. Indices (oc) and (sc) refer to open circuited and short circuited
electrodes, respectively. The convertible energy Uconv is the difference of the
energy for open circuited electrodes Uoc and short circuited electrodes Usc for a
given strain field Si. These energies will in general be represented by a volume
integral. For linear constitutive relations, the energy U for a piezoceramic solid
in the general form is represented by
U =1
2
∫V
[TiSi +DkEk] dV, (2.43)
3)Also, a particularly simple geometry and electrode configuration is required.
14 Chapter 2. Piezoelectric Ceramics
and an explicit expression is obtained after substitution of the dependent pair
of state variables by the independent pair via a set of constitutive equations e.g.
of the form (2.27) leading to
U =1
2
∫V
[cE11(S2
1 + S22) + cE33S
23 +
1
2(cE11 − cE12)S2
6 + 2cE12S1S2 (2.44)
+ 2cE13(S1 + S2)S3 + cE44(S24 + S2
5) + εS11(E21 + E2
2) + εS33E23
]dV,
where 2cE66 = cE11 − cE12, for symmetry reasons. Expression (2.44) is convenient
to compute the EMCF according to (2.42), as for a given strain distribution Si,
only the Ek depend on electrical boundary conditions. Still, depending on the
problem, other choices might be preferred.
The EMCF is a measure for the relative amount of energy that can be con-
verted from the mechanical to the electrical ports of the system and vice versa,
in a quasistatic deformation cycle. To illustrate the definition (2.42), an example
for the calculation of a material coupling factor, the longitudinal coupling factor
kl33 [IEE88], will be given.
3S
3T
scU Uconv0S
S0
S =3
b
a
c
c
b
a 3S =0
S03
S =
a
c b
Figure 2.1: quasistatic deformation cycle
A piezoelectric rod, initially stress free with electroded top and bottom sur-
faces, which are initially charge free, is subjected to a prescribed homogeneous
strain field S3 = S0 while the electrodes are open-circuited (free) and the rod is
free to cross expand, i.e. T1, T2 = 0 (fig. 2.1). From the constitutive equations
2.3. The Electromechanical Coupling Factor 15
(2.41), the stress T3 and electric field E3 corresponding to state b are
T3 = S0εT33
εT33sE33 − d233
, E3 = S0d33
d233 − εT33sE33, (2.45)
taking into account, that D3 = 0 at the electroded surfaces for charge free
electrodes and such for the entire volume4) of the piezoelectric rod. The energy
Uoc is computed as
Uoc =1
2V S2
0
εT33εT33s
E33 − d233
, (2.46)
E3 and T3 being the only non-vanishing variables in (2.43). V is the volume of
the rod.
The electrodes are then connected to an ideal electric load5), under constant
strain S0 and state c is reached. The work done on the ideal electric load is the
convertible energy of the system. Now, we have equal potential on the electrodes
and E3 = 0. For short-circuited electrodes, the stress T3 under given strain S0
becomes
T3 =S0
sE33, (2.47)
and the energy is
Usc =1
2V S2
0
1
sE33, (2.48)
where only T3 and S3=S0 contribute to the energy expression (2.43). To com-
plete the cycle, the strain field is removed, while the electrodes remain short-
circuited and the rod is back in its initial state a. The coupling factor (2.42) for
this deformation cycle becomes
(kl33)2 =
εT33εT33s
E33−d2
33− 1
sE33εT33
εT33sE33−d2
33
=d233εT33s
E33
, (2.49)
and the particularly simple form of this expression is a result of the homogeneous
one-dimensional deformation S0 and the simple geometry of the piezoelectric
rod. It should be noted here, that the result of this computation depends on
the condition T1, T2 = 0 for longitudinal expansion of the rod. The shaded area
4)According to (2.2), the electric displacement is divergence free, i.e. D3,3 = 0 inside thepiezoelectric body, where the density of free charges ρf is zero. So if D3 is continuous and zero
on the surface it has to be zero in the entire volume.5)This could be an ordinary resistor leading to an exponential decay of the potential differ-
ence on the electrodes. It is ideal in the sense that zero potential difference is reached in finitetime and the procedure still can be considered quasistatic.
16 Chapter 2. Piezoelectric Ceramics
in the T3, S3 diagram (fig. 2.1) represents the convertible energy Uconv of the
quasistatic deformation cycle.
kl33 is called a material coupling factor because the expression (2.49) contains
only constitutive parameters and therefore defines a material property. If the one
dimensional strain field S0 was not homogeneous, the expression (2.49) would
rather be a fraction of integral terms of the form (2.43), and the resulting effective
EMCF for inhomogeneous longitudinal deformation would always be smaller
than kl33. Also, the electrode geometry has an impact on the effective EMCF. So
in the general case of inhomogeneous deformation, the resulting coupling factor
calculated according to definition (2.42) would always be the effective coupling
factor. Only in the case of homogeneous fields, the resulting expression reduces
to a material coupling factor. For the terminology used in this work, the term
effective will be omitted, and the EMCF shall be understood as related to a
particular mode of deformation according to definition (2.42).
The definition (2.42) is convenient, because it relates to an arbitrary mode
of deformation of an electroded piezoelectric body, whereas material coupling
factors simply represent a set of constitutive parameters. Consequently, the
EMCF in this form can be used as a design criterion in actuator applications,
where the performance of the actuator as an energy transmitter does not only
depend on its material properties, but also on geometry, electrode shape and
deformation.
Clearly, in this definition no losses or dynamic effects are taken into account.
2.3.1 IEEE/IRE standards definitions
An early definition given in the IRE Standard on Piezoelectricity of 1958 [IRE58],
which is still occasionally found in the literature [LP91], [Ler90], is based on
energy expressions containing coupling terms similar to (2.30) which constitute
linear interaction between state variables. Accordingly, the EMCF is defined as
k2IRE =Um
2
UeUd, (2.50)
where the so-called mutual, elastic and dielectric energies are given by
Um = EidijkTjk, Ue = TijsEijklTkl, Ud = Eiε
TijEj , (2.51)
2.3. The Electromechanical Coupling Factor 17
respectively, with (2.30) as the energy expression. In this form, the EMCF can
be calculated per volume element and in the general case of inhomogeneous
field distributions, the EMCF for a given structure will be obtained by volume
integration.
In the case of simple homogeneous fields, (2.50) delivers the expected material
coupling factors introduced above. However, the definition (2.50) suffers from
at least one severe drawback. If the electroelastic state for a transducer can be
represented by one-dimensional fields, e.g. T1 and E3, kIRE reduces to a material
coupling factor independent of the deformation mode. For the given example,
with (2.30) as the energy function and T1, E3 as the independent variables, kIRE
becomesUm
2
UeUd=
(E3d31T1)2
T1sE11T1E3εT33E3=
d231sE11ε
T33
= k231,
where k31 is the material coupling factor for longitudinal deformation of a thick-
ness polarized piezoelectric rod [BCJ], which plays an important role in bending
actuation applications.
The above definition (2.50) was abandoned, and replaced by a new definition
in 1978 [IEE78] which is based on a quasi-static stress cycle. It remained un-
changed since then and in the ANSI/IEEE standard of 1987 [IEE88], which is
currently under revision, the calculation of the material EMCF kl33 is described
as follows (see fig. 2.2):
”The element is plated on faces perpendicular to x3, the polar axis, and is
short-circuited as a compressive stress −T3 is applied. The element is free tocross expand, so that T3 is the only nonzero stress component. From the figure it
can be seen, that the total stored energy per unit volume at maximum compression
is W1 + W2. Prior to removal of the compressive stress, the element is open-
circuited. It is then connected to an ideal electric load to complete the cycle. As
work is done on the electric load, the strain returns to its initial state. For the
idealized cycle illustrated, the work W1 done on the electric load and the part of
the energy unavailable to the electric load W2 are related to the coupling factor
kl33 as follows:”
(kl33)2 =W1
W1 +W2=sE33 − sD33sE33
=d233sE33ε
T33
(2.52)
This definition gives the same result for kl33 as (2.42) and could be extended to
nonhomogeneous fields with no extra effort. For nonlinear material properties,
18 Chapter 2. Piezoelectric Ceramics
slope s33E
slope s33D
3S
3T0T
W1
2W
T3
T0
= -
T3
T0
= -
a b
b c
c a
a
b
c
Figure 2.2: quasistatic stress cycle
the EMCF can be computed for a prescribed stress cycle [HPS+94] and the ex-
tension of the definition (2.42) to nonlinear constitutive relations would equally
be possible. The evaluation of the energy expressions will then generally require
numerical integration. However, the current standard [IEE88] limits the appli-
cation to linear material properties, homogeneous fields and material coupling
factors.
2.3.2 Dynamic/Effective EMCF
The definition (2.42) of the electromechanical coupling factor given above, refers
to a particular prescribed mode of deformation of the actuator. In transducer
applications, the mode and amplitude of vibration will depend on the frequency
of excitation and so will the performance of the actuator. On the other hand, the
frequency characteristics of an actuator can be used to estimate its electrome-
chanical coupling ability [BCJ]. For any conservative electromechanical system
vibrating freely in an eigenmode
ui (xk, t) = ui (xk) cos Ωt, (2.53)
φ (xk, t) = φ (xk) cos Ωt, (2.54)
2.3. The Electromechanical Coupling Factor 19
the maximum kinetic energy
Tmax =1
2Ω2
∫V
ρ uiui dV (2.55)
equals the maximum internal energy
Umax =1
2
∫V
(TiSi + DkEk) dV (2.56)
computed for the mode of vibration [ui, φ]. If the displacement field ui (xk) is
given, the electric potential field φ (xk) and also the internal energy Umax of the
system depend on the electrical boundary conditions. In particular, for open or
short-circuited electrodes the internal energies would be different for the same
displacement field.
In the case of forced vibration of a conservative system excited by a piezo-
electric actuator, resonance occurs when the frequency of excitation matches the
resonance frequency fr, which is the eigenfrequency of the same system with the
electrodes of the actuator short-circuited. In this case, the electrical impedance
of the system is zero. The eigenfrequencies of the system with free electrodes are
the so-called antiresonance frequencies fa. Being excited at these frequencies,
the system vibrates in the corresponding antiresonance modes and the electric
impedance of the system is infinite i.e. the amplitude of the electric current is
zero.
Both, the resonance and the antiresonance mode are eigenmodes of the sys-
tem, for short-circuited or open electrodes, respectively. Now, for the particular
(hypothetical) case where the displacement fields of resonance and antiresonance
are identical, the respective internal energies are computed for the same deforma-
tion but different electrical boundary conditions. Recalling the definition (2.42)
and having in mind, that (2.55) and (2.56) are equal, it follows that
k2d =Uoc − Usc
Uoc=
Ω2a − Ω2
r
Ω2a
=f2a − f2
r
f2a
(2.57)
for the case of identical resonance and antiresonance modes of defor-
mation. This is actually the case for the ballooning vibration mode of a thin
piezoelectric ring or spherical shell [BCJ], polarized in radial direction and fully
electroded on the inside and outside.
The frequency relation (2.57) is also referred to as the effective or dynamic
electromechanical coupling factor [Hun54], [BCJ] of a piezoelectric resonator.
20 Chapter 2. Piezoelectric Ceramics
Resonance and antiresonance frequencies fr, fa of a vibrating actuator can easily
be obtained by impedance measurements. Mason computed material coupling
factors for different piezoelectric materials from impedance measurements of lon-
gitudinally vibrating piezoelectric crystals [Mas50], but it was not his intention
to classify the performance of different transducers regarding energy conversion.
The relation (2.57) originated from equivalent circuit considerations. Repre-
senting a piezoelectric resonator by a simple equivalent circuit corresponds to a
modal discretization in one degree of freedom. It is no surprise, that the relation
(2.57) always holds for such an approximation, as resonance and antiresonance
modes are then necessarily identical.
The relation (2.57) is useful to compute approximate values of the EMCF of
an electroelastic system from impedance measurements which deliver the reso-
nance and antiresonance frequencies. These approximate values correspond to
those obtained from definition (2.42), if the prescribed strain field is the reso-
nance mode of vibration. Generally, the accuracy of this approximation will be
good if resonance and antiresonance modes are similar i.e. if the corresponding
frequencies are close and consequently the values for kd are small. Dynamic
effects are considered only in the sense that the underlying mode of deformation
is the resonance mode. Still, the cycle (fig. 2.1) that the structure is assumed
to undergo to determine the amount of energy converted, is a quasi-static hypo-
thetical process and not representative for the behavior of a dynamic system.
Chapter 3
Optimal Design
A useful optimization criterion to be employed in the design process of piezoce-
ramic actuators should give satisfying results for a variety of applications. On
the other hand, it should be very general and straightforward to use, with as
little as possible information required on the particular structure, the actuator is
going to be attached to. One crucial aspect will be the bonding of the actuator
to the structure, but information about e.g. the thickness of the bonding layer
might be difficult, if not impossible, to obtain. Of course, the energy transfer
into the structure will depend on its ability to emit energy itself, i.e. mechanical
boundary conditions. Again, these might be hard to determine, even if the com-
plete design of the structure is given. The structures under investigation here
are slender beams, i.e. the ratio of the longitudinal to the transverse dimensions
is large and they are assumed to deform according to the Bernoulli-Euler
hypothesis. The optimization variable will be the thickness of the piezoelectric
layer of the laminated beam. Under operating conditions, the performance of
the actuator will depend on the deformation and generally this deformation will
again depend on the thickness of the actuator. However, for analytical treatment
of the problem, the influence of changing actuator thickness on the beam vibra-
tions can only be considered for particularly simple geometries and boundary
conditions.
Models of the bending actuation of piezoelectric patches on beams and plates
have been presented by several authors [CR94]. Among them, the earliest so-
called Pin-Forcemodels for actuator pairs bonded symmetrically to both surfaces
22 Chapter 3. Optimal Design
1
3
electrode
PZT layer
substrate H
h
∆φ
e x
x
L
Figure 3.1: laminate beam structure
of a beam were introduced by Crawley, de Luis and Anderson [CdL87],
[CA90]. Pin-Force models neglect the added bending stiffness and mass ef-
fects due to the actuator patches. They are not suitable for the modeling of
thick piezoelectric layers on thin substrate beams [SBG95], [CR94]. Euler-
Bernoulli models assume a linear strain distribution in the piezoelectric layer
and include the added stiffness and mass of the actuator patches [BSW95]. For
actuators bonded to one side only, longitudinal and flexural vibrations are gen-
erally coupled [SBG95]. An assumption common to all the models is that the
electric field in the piezoelectric layer be independent of the thickness coordi-
nate, which does not comply with Gauss’s law (2.2). It will be shown that this
simplification is unnecessary and the correct linear distribution of the electric
field can be incorporated into the model without effort [GUS89].
3.1 The Laminate Beam
The portions of the beam to which actuators are bonded form a laminate, con-
sisting of the piezoceramic layer, which is electroded on the outward face, and
the substrate material (fig. 3.1). The substrate layer is assumed to be conduc-
tive and serves as the second electrode, such that a potential difference ∆φ can
be applied between the electrode and the substrate layer.
In compliance with the Bernoulli-Euler hypothesis, the cross sections of
the laminate remain plane and perpendicular to the neutral axis, which is located
3.1. The Laminate Beam 23
at a distance e to the interface layer and is inextensible, i.e. the laminate is
assumed to undergo pure bending deformation. The longitudinal displacement
u1 in the laminate is then
u1(x1, x3) = −x3w′(x1), (3.1)
and the longitudinal strain
S1 = u1,1 = −x3w′′(x1) (3.2)
follows from (2.10). The transverse stresses T2, T3 and shearing stresses T4, T5, T6are assumed to vanish as well as the shearing strains S4, S5, S6. Of the electrical
quantities, the only nonzero components are E3,D3, so we have 12 additional
equations:
T2, T3, T4, T5, T6, S4, S5, S6,D1,D2, E1, E2 = 0. (3.3)
Given these assumptions, the constitutive equations (2.41) reduce to
S1 = sE11T1 + d31E3, (3.4)
D3 = εT33E3 + d31T1, (3.5)
where the expressions for longitudinal strains S2, S3 have been omitted as these
do not contribute to the internal energy U (2.43) and therefore do not provide
any essential information. Substitution of (3.2) and (3.4) into (3.5) yields
D3 = (1 − k231)εT33E3 − d31sE11x3w
′′(x1) (3.6)
for the electric displacement. The expression (3.6) has been simplified by the
introduction of
k231 =d231sE11ε
T33
, (3.7)
which would be the EMCF of the piezoelectric layer (fig. 3.1) for uniform longi-
tudinal strain.1) From (2.2) it is seen, that the electric displacement is constant
over the thickness coordinate (D3,3 = 0) which in connection with (3.6) leads to
(1 − k231)εT33E3,3 =d31sE11w′′(x1), (3.8)
1)This is true for T2, T3 = 0. A uniform (homogeneous) longitudinal strain field correspondsto a uniform longitudinal extension, which does not comply with the strain field (3.2).
24 Chapter 3. Optimal Design
and the conclusion, that the electric field E3 is a linear function of the thickness
coordinate x3. Consequently, the potential φ which is related to the electric field
by (2.3) must be of the form
φ(x1, x3) = φ0(x1) + φ1(x1)x3 + φ2(x1)x23, (3.9)
so that
E3 = − ∂
∂x3[φ(x1, x3)] = −φ1(x1) − 2φ2(x1)x3 (3.10)
where the unknown functions φ1(x1), φ2(x1) have to be determined from (2.7)
and (3.8). The latter of these expressions leads to
φ2(x1) = −1
2
k231d31(1 − k231)
w′′(x1). (3.11)
An electric potential can only be defined relative to a reference and not as an
absolute value. Therefore, only the potential difference between the electrode
and the substrate
φ(x1, e+ h) − φ(x1, e) = ∆φ (3.12)
can be prescribed for the laminate (fig. 3.1). φ1 is now obtained by substitution
of (3.9) together with (3.11) into (3.12) and the resulting electric field
E3 =∆φ
h− 1
d31(1 − k231)
[(e+
h
2
)− x3
]w′′(x1) (3.13)
is uniquely determined by the potential difference ∆φ and the curvature w′′(x1)
of the laminate. Given the electric field, the longitudinal stress
T1 =1
sE11
[k231
(e+
h
2
)− x3
]w′′(x1) − d31
sE11
∆φ
h(3.14)
follows from (3.4). The internal energy expression (2.43) with (3.3) boils down
to
U =1
2
∫V
[sE11T
21 + εT33E
23 + 2d31E3T1
]dV, (3.15)
where E3, T1 are given by (3.13),(3.14) respectively. For given displacement
w(x1), (3.15) delivers the internal energy, depending on the potential difference
∆φ. For short circuited electrodes (∆φ = 0),
Usc =1
2
bh
sE11
[(e+
h
2
)2
+1
(1 − k231)
h2
12
]∫ L
0
[w′′(x1)]2 dx1 (3.16)
is obtained after integration over the volume V of the piezoelectric layer. To
complete the expression (2.42), still the internal energy must be computed for
3.2. Maximum EMCF 25
open circuited electrodes. The unknown potential difference between the elec-
trode and the substrate follows then from the integral boundary condition (2.36).
The electric displacement D3 yet has not been expressed as a function of ∆φ
and displacement w(x1). Substitution of (3.13) into (3.6) yields
D3 = (1 − k231)εT33∆φ
h+d31sE11
(e+
h
2
)w′′(x1), (3.17)
and presuming, that the electrode and substrate are initially charge free, (2.36)
turns into ∫ L
0
D3 dx1 = 0, (3.18)
which would have to be evaluated at the boundary x3 = e or x3 = e+h if D3 was
depending on the thickness coordinate. With the resulting potential difference
∆φoc =k231
d31(1 − k231)
h
L
(e+
h
2
)w′(x1)|L0 , (3.19)
the internal energy Uoc corresponding to state b (fig. 3.1) is found. Using
Uoc = Usc + Uconv, (3.20)
a relatively compact expression for the convertible energy
Uconv =1
2
bh
sE11
k231(1 − k231)
(e+
h
2
)21
L
[w′(x1)|L0
]2(3.21)
is obtained. The expression for Uoc is then simply the sum of (3.16) and (3.21).
3.2 Maximum EMCF
The definition (2.42) of the electromechanical coupling factor suggests, that only
contributions of piezoelectric material to the internal energy are to be accounted
for. However, it is by no means restricted to piezoelectric material. The laminate
structure under investigation is composed of the piezoelectric layer and the non
piezoelectric substrate layer. Given (3.3), the longitudinal stress in the substrate
becomes
T(s)1 = EsS1, (3.22)
26 Chapter 3. Optimal Design
where Es is Young’s modulus of the substrate material. The energy contribution
of the substrate layer
Us =1
2
∫Vs
S1T(s)1 dVs =
1
2bHEs
(e− H
2
)2 ∫ L
0
w′′(x1)2 dx1 (3.23)
is again obtained by integration over the corresponding volume Vs of the sub-
strate and Us is obviously independent of electrical boundary conditions.
So far, all of the energy expressions (3.16), (3.21) and (3.23) still depend on
the distance e of the laminate’s neutral axis to the interface layer. Demanding
that the normal force, i.e. the integral of the longitudinal stress over the cross-
sectional area, vanishes for the given pure bending deformation, the neutral axis
is determined from ∫ e
e−H
T(s)1 dx3 +
∫ e+h
e
T1 dx3 = 0, (3.24)
which, in connection with (3.22) and (3.14) results in
e =H
2
EssE11 − h2
H2
EssE11 + hH
(3.25)
for the location of the origin of the coordinate frame. Being not influenced by
electrical boundary conditions, (3.23) will clearly not contribute to the struc-
ture’s convertible energy, but rather to Uoc in the definition (2.42). The EMCF,
augmented by the contribution Us of non piezoelectric material is then given by
k2 =Uconv
Usc + Uconv + Us, (3.26)
a rather bulky expression, which is a functional of the prescribed deformation
w(x1). For a given mode of deformation, (3.26) becomes an explicit function of
the thickness ratio γ = hH of the layers only. However, (3.26) is an expression of
the form
k2(w, γ) =fconv(γ)
F [w(x1)][fsc(γ) + fs(γ)
]+ fconv(γ)
, (3.27)
where the only term depending on the prescribed deformation is the functional
F [w(x1)] =1[
w′(x1)|L0]2 ∫ L
0
[w′′(x1)]2
dx1. (3.28)
3.3. Maximum Dynamic EMCF 27
Now, assuming that the mode of deformation w(x1) and the thickness ratio γ
are not interrelated2), the derivative of (3.27) taken with respect to γ
d
dγ
[k2(w, γ)
]= F [w(x1)]
f ′conv(γ) [fs(γ) + fsc(γ)] − fconv(γ)[f ′s(γ) + f ′sc(γ)
]F [w(x1)]
[fsc(γ) + fs(γ)
]+ fconv(γ)
2(3.29)
shows, that extremal values of the EMCF, i.e. values of k2 for which
d
dγ
[k2(w, γ)
]= 0, (3.30)
can be found for values of the thickness ratio γ, which satisfy the condition
f ′conv(γ) [fs(γ) + fsc(γ)] − fconv(γ) [f ′s(γ) + f ′sc(γ)] = 0, (3.31)
independently of the deformation w(x1) of the laminate. Following the above
procedure, substituting the expressions (3.23), (3.21), (3.16) with (3.25) into
(3.26) and taking the derivative with respect to γ, the condition (3.31) for ex-
tremal values can be written in the form
µ2(1 + γ)(γ2 + 2γ + µ)(2γ4 + 5µγ3 + 3µγ2 − 3µ2γ − µ2) = 0, (3.32)
showing that only the ratio of the longitudinal stiffnesses of the layers µ = EssE11,
and not piezoelectric or dielectric constants, determine the thickness ratio for
maximum electromechanical coupling according to definition (2.42). Computa-
tional results for the EMCF depending on the thickness ratio are given in chapter
four.
3.3 Maximum Dynamic EMCF
The optimal thickness for maximum EMCF based on definition (2.42) of the
laminate beam was independent of the assumed deformation. For a vibrating
electromechanical system, the changing actuator thickness will generally have an
impact on the mode of vibration. For the particular case of a periodic laminate
2)Generally, the mode of vibration of a piezoelectric laminate will depend on the thicknessratio of the two layers. This dependence can only be modeled, if explicit solutions to the
equations of motion are available. The EMCF is computed for a prescribed mode of defor-mation here, and a possible influence of the thickness ratio on this mode is neglected for theoptimization.
28 Chapter 3. Optimal Design
beam (fig. 3.2), explicit solutions for the equations of motion can be found
and the dynamic EMCF can be computed according to definition (2.57). The
dashed lines mark the locations where periodic boundary conditions apply. The
piezoelectric bottom layer is polarized in thickness direction with alternating
orientation corresponding to the distinctly shaded areas. A harmonic voltage
∆φ = U0 sin(Ωt) is applied between the bottom electrode and the substrate
layer.
1
∆φ
x
L L
(1) (2) H
h
Figure 3.2: periodic laminate beam system
For the periodic laminate beam, given the assumptions (3.2), (3.3), the gen-
eral form (2.12) of the electric enthalpy density reduces to
H =1
2(T1S1 −D3E3) (3.33)
for linear constitutive relations (3.4),(3.5), which upon substitution of (3.2),
(3.6), (3.13), (3.14) and integration over the thickness h of the layer leads to
Hp =1
2
bh
sE11
∫ L
0
[(∆φ
h
)2
+ 2d31
(e+
h
2
)∆φ
hw′′(x1)
+
[(e+
h
2
)2
+1
1 − k231h2
12
][w′′(x1)]
2
]dx1 (3.34)
for the electric enthalpy of the piezoelectric layer. For the substrate layer, the
quantities E3 and D3 vanish, such that internal energy density U and electric
enthalpy density H are equivalent here (2.12). Consequently, (3.23) represents
the contribution of the substrate layer to the electric enthalpy of the laminate
and with the corresponding Lagrangeian (2.16), Hamilton’s principle for the
laminate is written as∫ t2
t1
[∫ L
0
1
2ρbh δw2 dx1 − δ(Hp + Us)
]dt+
∫ t2
t1
∫S
(σδφ− f δui) dS dt = 0,
(3.35)
3.3. Maximum Dynamic EMCF 29
and the displacement w(x1, t) is now time dependent. After partial integration,
the expression∫ L
0
(CwIV + ρbhw)δw dx1 − Cw′′′δw|L0 + [Cw′′ +M0] δw′|L0 = 0 (3.36)
is obtained, where the effective bending stiffness
C =bh
sE11
[(e+
h
2
)2
+1
1 − k231h2
12
]+ bHEs
[(e− H
2
)2
+H2
12
](3.37)
of the laminate and the actuation bending moment
M0(t) = bd31sE11
(e+
h
2
)∆φ(t)
have been introduced.
For forced vibrations induced by the harmonic voltage ∆φ(t) = U0 sin(Ωt),
solutions are found in the form
w(x1, t) =1
2
M0
CK2
[cosK(x1 + L
2 )
cosK L2
− coshK(x1 + L2 )
coshK L2
](3.38)
where the wave number K is defined by the dispersion relation
K4 = Ω2 ρb(h+H)
C(3.39)
[Fre97], [WFHS97]. Resonant modes of vibration of the laminate are repre-
sented by wave numbers K which satisfy
cosKL
2= 0,
and the corresponding resonance frequencies are then obtained from the disper-
sion relation (3.39).
The charge on the bottom electrode is determined by (2.36) and (3.17) leading
to
Qs(t) = 2bL
[εT33
∆φ(t)
h(1 − k231) +
d31sE11
(e+h
2)[w′(t)|L0 − w′(t)|0−L
] 1
2L
](3.40)
30 Chapter 3. Optimal Design
and the time derivative of this expression gives the current through the consid-
ered part of the periodic system. Upon substitution of the solution (3.38) into
(3.40) and taking the time derivative, the condition for zero current
tanKL
2+ tanh
KL
2+KL
CsE11bh(e+ h
2 )2(1 − k231)
k231= 0 (3.41)
is obtained. Wave numbers that satisfy this expression have to be determined
numerically. They represent the antiresonance modes of the laminate and the
corresponding antiresonance frequencies fa = Ωa/2π result again from the dis-
persion relation (3.39). The dynamic EMCF
k2d =f2a − f2
r
f2a
(3.42)
is numerically computed for the periodic system (fig. 3.2) for different thickness
ratios H/h of the layers. Results are given in chapter four.
3.4 Maximum Power Factor
The EMCF is a measure for the fraction of the maximal work supplied to a
electromechanical system, which is converted in a quasi-static deformation cycle.
This cycle has hypothetical character and is convenient for computation, but
dynamic load cycles will generally be different. The consideration of dynamic
cycles in this work is restricted to steady state vibrations of linear electroelastic
systems. Then, the load cycles are ellipses in the (Ti, Si) and (Ek,Dk) state
planes and it is convenient to represent the state variables in complex notation
Si(xj , t) = Si(xj) cos[ωt+ φSi(xj)]
= Si(xj)e[jωt+φSi(xj)]
= Si(xj)ejωt, (3.43)
where Si(xj) is the real amplitude and
Si(xj) = Si(xj)ejφSi
(xj) (3.44)
the complex amplitude, including the phase angle φSi . A corresponding nota-
tion could be adopted for the Ti, Ek and Dk, but for simplicity the extra hats
3.4. Maximum Power Factor 31
and underlines shall be omitted and all time dependent quantities will be substi-
tuted by their complex amplitudes. In this complex representation, Poynting’s
theorem for a piezoelectric body is found to be [Aul89]∫S
Pini dS = jω1
2
∫V
(TiS∗i +EkD
∗k − ρviv∗i ) dV − 1
2
∫V
Fiv∗i dV, (3.45)
where Pi is the complex Poynting vector
Pi =1
2
[φ [jωDi]
∗ + P(m)i
](3.46)
and ni the outward unit normal vector on the surface of the body. The in-
tegrations are carried out over the surface S and the volume V of the body,
the asterisk denotes the conjugate of the complex amplitudes. Elastic contribu-
tions P(m)i to the complex power flow cannot be represented in the engineering
notation used here. In standard tensor notation they are P(m)i = −Tijv∗j .
The integral terms on the right hand side of (3.45) can be identified as
Upeak =1
2
∫V
(TpS∗p +EkD
∗k) dV,
Tpeak =1
2
∫V
ρviv∗i dV, (3.47)
the peak potential and kinetic energy of the system, respectively. Obviously,
Tpeak is a real expression and Upeak is also real, provided the constitutive relations
satisfy [SpDk
]∗=
[sEpqdkq
dplεTkl
] [TqEl
]∗, (3.48)
which is true, if the matrix of the constitutive coefficients is real and symmetric.
If the body is free of volume forces Fi, Poynting’s theorem can be written as
1
2
∫S
[φ [jωDi]
∗ + P(m)i
]ni dS = jω(Upeak − Tpeak), (3.49)
and it is seen, that expressions on both sides are imaginary quantities.
Energy conversion in lossless piezoelectric media is represented by (3.49) as
the real part of the power flow Pini across the surface of the considered volume.
If it is assumed, that stresses Tij and surface charge density σ = −Dini are
present on distinct surfaces ST and Sσ only, the real part of (3.49) becomes
1
2∫
Sσ
φ [jωDi]∗ni dS
+
1
2∫
ST
P(m)i ni dS
= 0, (3.50)
32 Chapter 3. Optimal Design
and if ST and Sσ are identified as the mechanical and electrical ports of the
system, this simply states that resistive power flow through the mechanical port
equals minus the resistive power flow through the electrical port.
In the following analysis, energy conversion from the electrical to the me-
chanical port of the system shall be considered. The piezoelectric laminate (fig.
3.1) is excited by a harmonic voltage ∆φ = V (t) = V0ejΩt. At the electrical
port, the complex power input is then
Pel =1
2
∫Sσ
φ [jωDi]∗ni dS (3.51)
where the integration has to be carried out over the electroded top and bottom
surfaces corresponding to Sσ. The dielectric displacement Di is given by (3.17)
leading to
Pel =1
2V0jΩbL
[(1 − k231)εT33
V0h
+d31sE11
(e+h
2)
1
Lw′|L0]∗
=1
2V0I
∗0 , (3.52)
where the factor multiplied by 12V0 can be recognized as the conjugate of the
complex amplitude I0 of the current
I(t) = I0ejΩt = −jΩbL
[(1 − k231)εT33
V0h
+d31sE11
(e+h
2)
1
Lw′|L0]ejΩt (3.53)
which depends on the applied voltage V0 and the resulting displacement w(x1)
of the laminate. The work Wconv that is converted from the electrical to the
mechanical port per harmonic deformation cycle is proportional to the resistive
power, which is the real part of the electric power, according to
Wconv =2π
ΩPel, (3.54)
and obviously this expression corresponds to Uconv in the definition of the EMCF
(2.42) which is based on a quasi-static cycle of deformation and discharge. Now,
to come up with a similar definition for harmonic cycles, a corresponding ex-
pression for Uoc or in Mason’s words ”the total input electrical energy obtained
from the input battery” must be found. For harmonic cycles, where energy trans-
fer is continuous rather than sequential as in the case of the quasi-static cycle,
the maximum of the work supplied in one cycle is dependent on the choice of
the start/end point of the cycle. This is a typical problem encountered in the
treatment of alternating or arbitrary time dependent currents and voltages and
3.4. Maximum Power Factor 33
it is customary to represent these time dependent quantities by their effective
values, defined as
I2 =1
T
∫ T
0
I2(t) dt, (3.55)
where T is typically the period, but can also go to infinity for nonperiodic signals.
The apparent power is the product of the effective voltage and current so that
for harmonic signals
V I =1
2|V0I∗0 | = |Pel| (3.56)
and the work expression resulting from the apparent power will be considered
equivalent to the maximum work supplied from the electric power source in one
cycle, which is clearly different from the converted work (3.54). Rather than
taking the ratio of the time averaged converted and maximum work, the ratio
of the corresponding power expressions
Pel|Pel| =
V0I∗0|V0I∗0 |
= cosφ, (3.57)
is taken, where φ is the phase lag between voltage and current. This ratio, which
is also termed the power factor, is consequently the equivalent to the EMCF
for harmonic stationary load cycles. In some technical applications, reactive
currents can become quite large compared to resistive currents, which are in
phase with the applied voltage, leading to small power factors. This is undesired
because the resistive losses are proportional to the square of the current and
efforts are made to compensate reactive contributions by added capacitors or
inductors to maximize the power factor.
The current (3.53) can be separated into two distinct contributions. The
dielectric current
Id = −jΩbLεT33V0h
(1 − k231), (3.58)
is independent of the deformation of the body, corresponding to the capacitive
behavior of a clamped (Si = 0) piezoelectric actuator and always purely reactive
if the piezoelectric layer is free of dielectric losses. It does not contribute to the
resistive power. The piezoelectric current
Ip = −jΩbLd31sE11
(e+h
2)
1
Lw′|L0 , (3.59)
will depend on the unknown complex rotation amplitude w′(x1) of the cross-
sections at the boundary x1 = 0 and x1 = L of the laminate, which is propor-
tional to the applied voltage V0. The piezoelectric current can be expressed as
34 Chapter 3. Optimal Design
a function of1
Lw′|L0 = V0Ψ, (3.60)
where Ψ is a complex valued function of the frequency of excitation Ω. Without
loss of generality, V0 can be assumed real, and the power factor (3.57) is en-
tirely determined by the complex current amplitude I0. Provided, the resistive
power Pel is positive, finding a maximum of the power factor is equivalent
to minimizing the expression[PelPel
]2=
1
Ψ2[d31(1 − k231)
k231(e+ h2 )h
+ Ψ]2
(3.61)
and further analysis of this expression is not promising without consideration of
the influence of the thickness h of the piezoelectric layer on the complex steady
state response w(x1) of the forced vibration problem. Again, explicit solutions
for the equations of motion can only be found for very simple configurations. For
reasonably complex systems, the optimal geometry of a piezoelectric actuator
can be determined based on numerical solutions. This is demonstrated in chapter
four.
The concept of the determination of the optimal actuator design and location
based on the actuator power factor was also introduced by Liang, Sun, Zhou
and Rogers [ZR95], [LSR95]. They justified the use of the actuator power
factor, defined as the ratio
ψ =Dissipative Mechanical Power
Supplied Electrical Power=
Ys|Ys| , (3.62)
where Ys is the electrical admittance of the actuator, by general impedance
considerations [LSR97] rather than the demonstrated analogy to the EMCF.
Chapter 4
Verification
4.1 EMCF: Theory versus Finite Element Model for a pe-riodic system
There is a variety of finite element codes capable of modeling coupled field in-
teractions, including the piezoelectric effect. The ANSYS finite element package
has been used to compute resonance and antiresonance frequencies fr, fa of the
periodic laminate beam (fig. 3.2) for different thickness ratios H/h. The ge-
ometry of the beam is discretized into quadrilateral 2D elements (fig. 4.1).
Displacement degrees of freedom are coupled at the periodic boundaries, voltage
degrees of freedom are coupled along the electrode and the boundary to the non
piezoelectric substrate layer, respectively.
Figure 4.1: finite element mesh for the periodic laminate beam
A modal analysis has been carried out for resonance (fixed zero potential)
and antiresonance (free potential) electric boundary conditions, which are ap-
plied on the electrode and the boundary between piezoelectric and substrate
layer. From the computed eigenfrequencies, the dynamic EMCF (2.57) was cal-
culated and plotted over the thickness ratio h/H of the laminate beam for two
different substrate materials and two different aspect ratios L/h (figs. 4.2 and
36 Chapter 4. Verification
0 1 2 3 40
0.02
0.04
0.06
0.08
0.1
0.12
k2
kd2 beam theory
kd2 FEM: large aspect ratio
kd2 FEM: small aspect ratio
H/h
EM
CF
Figure 4.2: EMCF over thickness ratio H/h; substrate: aluminum
4.3). The solid curves show the EMCF according to definition (2.42) determined
analytically for a prescribed harmonic (sinusoidal) mode of deformation of the
laminate. Circles represent values of the dynamic EMCF calculated from eqns.
(3.41) and (3.39). Stars and diamonds are finite element results of the dynamic
EMCF for large and small aspect ratio.
The available 2D elements are restricted to plane strain models so the mate-
rial constants used for analytical computation of the EMCF and dynamic EMCF
for the considered uniaxial stress in the Bernoulli-Euler laminate had to be
adapted to give plane strain results. Based on the representation used in the
finite element package, where (Si, Ei) are independent variables, taking into
consideration that the laminate is free of shear deformation and of the elec-
trical quantities, only the ones in the direction of polarization are of interest,
constitutive equations in the form
T1 = cE11S1 + cE12S2 + cE13S3 − e31E3,
T2 = cE12S1 + cE11S2 + cE13S3 − e31E3,
T3 = cE13(S1 + S2) + cE33S3 − e33E3,
D3 = e31(S1 + S2) + e33S3 + εS33E3, (4.1)
are obtained for the Bernoulli-Euler laminate. In addition, the strain S2 in
4.1. EMCF: Theory versus Finite Element Model for a periodic system 37
0 1 2 3 40
0.02
0.04
0.06
0.08
0.1
0.12
k2
kd2 beam theory
kd2 FEM: large aspect ratio
kd2 FEM: small aspect ratio
H/h
EM
CF
Figure 4.3: EMCF over thickness ratio H/h; substrate: steel
width direction and the stress T3 in thickness direction vanish. Solving (4.1) for
(S1,D3) leads to constitutive equations
S1 =1
cE11cE33 − cE13cE13
[cE33T1 −
(cE13e33 − cE33e31
)E3
],
D3 =1
cE11cE33 − cE13cE13
[(cE13e33 − cE33e31)T1
+[cE13c
E13ε
S33 + 2cE13e31e33 − cE33(e231 + cE11ε
S33) − cE11e233
]E3
], (4.2)
which are of the form (3.4), (3.5). These are suitable to obtain analytical results
from the expressions derived for the laminate in chapter three to be compared
with plane strain finite element computations.
As it is evident from table 4.1, the computed analytical and numerical re-
sults for resonance and antiresonance frequencies are very close. Still, there is
a noticeable difference in the values obtained for the dynamic EMCF, as the
frequency based formula (3.42) is very sensitive to frequency deviations.
Figure (4.4) shows an example of resonance and antiresonance modes of vi-
bration of the bulky periodic laminate beam, for a thickness ratio H/h of two.
38 Chapter 4. Verification
Analysis FEM
H/h fr [Hz] fa [Hz] k2d fr [Hz] fa [Hz] k2d2.0 598.548 625.559 0.0844 598.088 624.333 0.0823
2.2 643.589 671.268 0.0807 643.019 669.974 0.0788
2.4 689.236 717.468 0.0771 688.525 716.069 0.0755
2.6 735.397 764.091 0.0736 734.515 762.552 0.0722
2.8 781.997 811.079 0.0704 780.908 809.360 0.0691
3.0 828.972 858.383 0.0673 827.643 856.447 0.0661
Table 4.1: resonance and antiresonance frequencies for the periodic laminate
beam
The deformation is very similar, but the electric potential distribution, repre-
sented by isopotential lines, in the piezoelectric layer is quite different for the two
modes. For the periodic Bernoulli-Euler laminate, the piezoelectric current
(3.59) is purely reactive as no energy is dissipated or transformed and is either
in phase or 180 out of phase with the dielectric current (3.58). Antiresonance
frequencies are generally higher and for this system only little higher than the
corresponding resonance frequencies.
When piezoelectric and dielectric current are in phase slightly below reso-
nance, they will be opposite in sign slightly above resonance. Under antires-
onance conditions these two contributions compensate each other and it is no
surprise that resonance and antiresonance modes can be so similar. It is not
the case, that the particular strain distribution of the antiresonance mode would
lead to zero piezoelectric current.
MN
MX
MN
MX
Figure 4.4: resonance (top) and antiresonance (bottom) eigenmode of the lami-
nated periodic beam; isopotential lines represent the electric potential distribu-
tion in the piezoelectric layer
4.2. Power Factor: Theory versus Experiment for an infinite beam 39
4.2 Power Factor: Theory versus Experiment for an infi-nite beam
In the previous chapter, analytical investigations aimed at finding an optimal
thickness of the piezoelectric layer without solving the equations of motion for
a particular actuator application. This has been partially successful in the case
of the EMCF, where optimum thickness ratios could be given based on a hypo-
thetical prescribed cycle of deformation. For the power factor as optimization
variable this was not possible in this general way and in this section optimal
dimensions of piezoelectric bending actuators shall be determined based on nu-
merical solutions of the equations of motion. To verify the predictions of the
mathematical model, numerical results are compared to measured data obtained
from an experimental setup [Sto98].
4.2.1 The problem geometry
Continuous transfer of resistive power into an electromechanical system requires
either some kind of dissipation mechanism or boundary conditions, which allow
for energy transfer across the boundaries of the system. The geometry here con-
sidered is a piezoelectric patch bonded to an infinite beam (fig. 4.5), which was
also suggested by Gibbs and Fuller, who investigated simultaneous actuation
of flexural and longitudinal waves for actuator pairs on infinite beams [GF92].
x1
x3
h
H
s
H/2
∆φL/2 L/2
Figure 4.5: piezoelectric patch bonded to infinite beam
4.2.2 Model for the bonded actuator
For the kinematical model of the single bonded actuator (fig. 4.6), the piezoelec-
tric patch and the substrate beam are assumed to undergo bending as well as
40 Chapter 4. Verification
longitudinal deformation. The bonding layer is shear elastic but rigid in thick-
ness direction and therefore allows for relative longitudinal displacement of the
bonded surfaces of the actuator and substrate layer. A kinematic relation for
x3
x1
,w
,u-
γ
+N N
M- +M
Q+-Q
Figure 4.6: laminate section with boundary forces/moments
the displacement corresponding to these assumptions is obtained as
us(x1, x3, t) = u(x1, t) − x3w′(x1, t),
up(x1, x3, t) = u(x1, t) − x3w′(x1, t) + sγ(x1, t), (4.3)
where the two dimensional fields us(x1, x3, t), up(x1, x3, t) of the displacement in
longitudinal (x1) direction for the substrate and the piezoelectric layer, respec-
tively, are represented by the one dimensional fields w(x1, t), for the transverse
displacement of the entire laminate in x3 direction, u(x1, t) for the longitudinal
displacement of the middle layer of the substrate beam and γ(x1, t), the shear
angle of the thin bonding layer, which is assumed constant over the thickness s
of the layer.
Based on the kinematic relation (4.3), potential and kinetic energies of the
laminate section can be expressed in terms of the independent variables u(x1, t), w(x1, t)
and γ(x1, t). For the kinetic energy density Ts(x1, t) of the substrate, the ex-
pression
Ts(x1, t) =1
2ρs[u2(x1, t) + w2(x1, t)
](4.4)
is obtained where terms proportional to w′, which represent the rotational energy
of the cross-sections of the beam and patch are neglected. The potential energy
density becomes
Us(x1, x3, t) =1
2Es u
′2s (x1, x3, t). (4.5)
As the mass density of the bonding layer is small compared to the other layers,
4.2. Power Factor: Theory versus Experiment for an infinite beam 41
its kinetic energy contribution is neglected and only the potential energy density
Ub(x1, t) =1
2Gb γ
2(x1, t), (4.6)
stored in shear deformation is taken into account. In the piezoelectric patch, the
kinetic energy density
Tp(x1, t) =1
2ρp[(u(x1, t) + s γ(x1, t))
2 + w2(x1, t)]
(4.7)
contains additional terms in γ due to the shear elasticity of the bonding layer.
Again, as for the substrate layer, rotational energy terms in w′ are neglected.
With the constitutive equations (3.4), (3.5) for the piezoelectric layer and
the displacement field (4.3), following the derivation given in chapter three, the
electric field
E3(x1, x3, t) =k231
d31(1 − k231)
[x3 − h+ 2s+H
2
]w′′(x1, t) − ∆φ(t)
h(4.8)
is again a linear function of the thickness coordinate x3 and proportional to both,
the applied voltage ∆φ(t) and the curvature w′′(x1, t) of the laminate section.
The electric enthalpy densityHp of the piezoelectric layer corresponding to (3.33)
can now be expressed in displacement (u,w, γ) and electric potential φ as the
independent quantities, however, the resulting expression is rather bulky and
will not be given here. The resulting Lagrangeian of the laminate section
L =
∫V
(Tp + Ts −Hp − Us − Ub) dV
with displacements and electric potential as the independent variables is suitable
for substitution into (2.18).
For a state change δu, δw of the laminate section, the virtual work of the
external forces done on the laminate is
δW = Q+δw+ +N+δu+ −M+δw′+ −N−δu− −Q−δw− +M−δw′
− , (4.9)
where the bending moment M , the shear force Q, the normal force N and the
corresponding rotations and displacements w′, w and u in the substrate beam are
evaluated at the boundaries of the laminate section in positive (+) and negative
(−) x1 direction, which is indicated by the subscripts. The virtual work of the
external charges in expression (2.17) is zero as for the given prescribed potential
boundary conditions no arbitrary variation of the potential on the electrodes is
permitted. Consequently, (4.9) represents the total virtual work done on the
42 Chapter 4. Verification
laminate section and can be substituted together with the Lagrangeian (4.2.2)
into Hamilton’s principle (2.18). After partial integration and collection of
terms in the same independent variations, the equations of motion of the system
are obtained. So far, the forces N,Q and bending moment M at the boundaries
of the laminate section are still unknown.
A single piezoelectric patch on an infinite beam can be thought of as a lam-
inate section (fig. 4.6) connected to two semi-infinite beams. For the steady
state response to harmonic voltage excitation ∆φ = V0 cos Ωt, force and bend-
ing moment amplitudes at the boundary of the laminate section can then be
determined in complex representation from a dynamic stiffness matrix for the
semi-infinite beam [CH88] according to Q+
M+
N+
=
b11+ b12+ 0
b21+ b22+ 0
0 0 b33+
w+
w′+
u+
, (4.10)
where the complex coefficients bij+ are frequency dependent. For the semi-
infinite beam, the bij+ become
b11+ = (−1 + j)EIk3w,
b12+ = −jEIk2w,b21+ = −b12+,b22+ = (1 + j)EIkw,
b33+ = jΩA√ρE, (4.11)
where E is Young’s modulus, I the moment of inertia of the cross-section and
A the cross-sectional area of the substrate beam. For the second boundary,
the coefficients bij− may differ only in the sign, depending on the choice of the
coordinate frame xi and the orientation of forces Q,N and bending moment M .
The dispersion relation
k4w = Ω2 ρA
EI
for flexural waves relates the wavenumber kw to the frequency of vibration, with
ρ as the mass density of the substrate beam.
The steady state response of a finite beam excited by a piezoelectric patch
can be modeled by connecting finite length beam ends to the laminate section.
For a finite beam of length ls, the corresponding bij+ of the dynamic stiffness
matrix are
B = 1 + e2 j kw ls + 4 e(1+j)kw ls + e2 kw ls + e(2+2 j) kw ls ,
4.2. Power Factor: Theory versus Experiment for an infinite beam 43
b11+ =1
B(1 + j)(−j + e2 j kw ls − e2 kw ls + j e(2+2 j)kw ls)EI k3w,
b12+ =1
Bj(−1 + e2 j kw ls)(−1 + e2 kw ls)EI k2w,
b21+ = −b12+,b22+ =
1
B(1 + j)(−1 + j e2 j kw ls − j e2 kw ls + e(2+2 j)kw ls)EI kw,
b33+ = −ΩA√ρE tan
[Ωls
√ρ
E
], (4.12)
and for the bij− again, only the sign may be different.
In either case, for semi-infinite or finite beam ends, the unknown Q,M,N
at the boundaries of the laminate section can be expressed by the independent
displacement quantities w, u and after separation of time, according to
γ(x1, t) = Γ(x1)ejΩt,u(x1, t) = U(x1) ejΩt,w(x1, t) = W (x1) ejΩt, (4.13)
a system of three coupled equations of motion in the form
c1 Γ + c5 Ω2 U + c6 Ω2 Γ − c11 U ′ − c12 Γ′′ − c13W ′′′ = 0,
c2 Ω2 U + c3 Ω2 Γ − c7 U ′′ − c8 Γ′′ − c9W ′′′ = 0, (4.14)
c4 Ω2W + c15 U′′′ + c16 Γ′′′ + c17W
IV = 0,
with inhomogeneous boundary conditions
−c7 U ′ − c8Γ′ − c9W ′′ + b33 U = c10 V0−c11 U ′ − c12 Γ′ − c13W ′′ = c14 V0
−c15 U ′ − c16 Γ′ − c17W ′′ + b21W + b22W′ = c18 V0
c15 U′′ + c16 Γ′′ + c17W
′′′ + b11W + b12W′ = 0
∣∣∣∣∣∣∣∣x1=+L
2 ,
(4.15)
−c7 U ′ − c8Γ′ − c9W ′′ − b33 U = c10 V0−c11 U ′ − c12 Γ′ − c13W ′′ = c14 V0
−c15 U ′ − c16 Γ′ − c17W ′′ + b21W − b22W ′ = c18 V0c15 U
′′ + c16 Γ′′ + c17W′′′ − b11W + b12W
′ = 0
∣∣∣∣∣∣∣∣x1=−L
2 ,
is found. The constants c1−18
44 Chapter 4. Verification
c1 = −bGK s,
c2 = b(h ρP +H ρS),
c3 = b s h ρP ,
c4 = b(h ρP +H ρS),
c5 = b s h ρP ,
c6 = b s2 h ρP ,
c7 = −b (H ES + h/sE11),
c8 = −b s h/sE11,c9 = b h
h+H + 2 s
2 sE11,
c10 = −b d31/sE11,c11 = −b s h/sE11,c12 = c11 s,
c13 = b s hh+H + 2 s
2 sE11,
c14 = −b s d31/sE11,c15 = h b
h+H + 2 s
2 sE11,
c16 = s c15,
c17 = −b(h3(−4 + 3 k231) + 3h(−1 + k231)(H + 2 s)(2h+H + 2 s) +
ESH3 (−1 + k231) sE11)/(12 (−1 + k231) sE11),
c18 = bd31(2 e+ h)
2 sE11(4.16)
for the infinite beam contain material and geometrical parameters of the system.
For the solution of the coupled system (4.14) in U , Γ and W the ansatz
U(x1) = U eλx1 ,
Γ(x1) = Γ eλx1 ,
W (x1) = W eλx1 (4.17)
is chosen, leading to an algebraic system of equations which is represented in
matrix form
M
U
Γ
W
= 0, (4.18)
4.2. Power Factor: Theory versus Experiment for an infinite beam 45
with
M =
c5 Ω2 − c11 λ2 −c13 λ3 c1 + c6 Ω2 − c12 λ2c2 Ω2 − c7 λ2 −c9 λ3 c3 Ω2 − c8 λ2c15 λ
3 c4 Ω2 + c17 λ4 c16 λ
3
.
(4.19)
For a nontrivial solution of the eqns. (4.18) the determinant of M must
vanish
detM = 0 (4.20)
and this characteristic equation is satisfied by each of the eight eigenvalues
λi, (i = 1...8), which are computed numerically using MATLAB. The corre-
sponding eigenvectors
Li =
U i
Γi
1
(4.21)
are determined from (4.18) by solving the system[1st row ofM (λi)
2nd row ofM(λi)
]Li = 0, (4.22)
where it is assumed that the first two rows of M(λi) are linearly independent1).
The general solution
LH =
8∑i=1
Ci Li eλi x1 (4.23)
of the homogeneous system (4.14) is a linear combination of vectors Li. From
the inhomogeneous boundary conditions (4.15) the linear coefficients Ci are de-
termined. These give the particular solution
L* =
U*
Γ*
W *
(4.24)
of the homogeneous system (4.14) with
U* =
8∑i=1
Ci U i eλi x, (4.25)
Γ* =
8∑i=1
Ci Γi eλi x, (4.26)
W * =8∑
i=1
Ci eλi x. (4.27)
1)If this was not the case, any other two rows ofM(λi) could serve to obtain the eigenvectors.
46 Chapter 4. Verification
These solutions represent the steady-state response of the actuator, from
which the input electric admittance, the power factor and also the displacement
of the finite or infinite beam can be computed. Due to the non-symmetrically
mounted patch, longitudinal and flexural waves are coupled.
p
PZTI
V
V
Cb
0
Figure 4.7: capacity Cb of the bonding layer
In other investigations of actuator patches bonded to beams, where the same
patch geometry and adhesive was used [Ngu99], the bonding layer was identified
ideally elastic. Consequently it is modeled mechanically lossless here. Dielectric
properties of the bonding material come into play as only the bottom electrode
of the piezoelectric patch is connected to the power source. The conducting sub-
strate layer serves as the second electrode and the potential on the free electrode,
which is bonded to the substrate layer, depends on the capacitive influence of
the bonding layer, having capacity Cb. Given the current I (fig. 4.7) the true
potential difference Vp across the electrodes of the PZT patch can be computed
according to
Vp =I
IV0
+ 1jΩCb
(4.28)
where V0 is the potential difference between bottom electrode and substrate
layer.
Material loss in the actuator patch is taken into consideration here by the
employment of complex values for the dielectric permittivity εT33 and the com-
pliance sE11. Real and imaginary parts
sE11′ − j sE11
′′= sE11(1 − j
Q)
εT33′ − j εT33
′′= εT33(1 − j tan δ) (4.29)
of these material parameters result from the dielectric loss tan δ and the quality
factor Q which are provided by the manufacturer of the piezoelectric patches.
Complex values for the piezoelectric constant d31 can be identified in ex-
periments but they are not independent of sE11′′
and εT33′′, and have to satisfy
restrictions, to make sure no energy would be generated in the material [HN69].
4.2. Power Factor: Theory versus Experiment for an infinite beam 47
0 5 10 15 200
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
f[kHz]
P [W
]
Figure 4.8: energy balance of the actuator patch on infinite beam (simulation): (−·−)
resistive power input (electrical); (—) resistive power output (mechanical: longitudinal
and flexural contributions); (−−) resistive power output (mechanical: only flexural
waves); h=1mm, L=70mm, b=25mm, H=3mm, s=0, material: PIC 141 on steel
For this reason, and also because only Q and tan δ were given for the piezoelec-
tric material used in the experiments, only real values for d31 are employed. It
turned out, that for the simulated single patch on an infinite beam, the power
loss in the piezoelectric patch is in the same order of magnitude as the power
that is actually converted into longitudinal and flexural waves (fig. 4.8).
4.2.3 Experimental setup
In an experimental setup, the energy transfer of single piezoelectric actuators
bonded to steel beams was tested. The actuators were positioned in the center
while the ends of the steel beam were embedded in sand (fig. 4.9) to avoid the
reflection of flexural and also longitudinal waves, originating from the laminate
region in the center of the beam. About three quarters of the beam were covered
with sand, while the middle section was free.
Beams of different thicknesses (1mm, 3mm, 5mm) were tested, each approx-
imately three meters in length and 25mm wide. Rectangular 70mm×25mm
48 Chapter 4. Verification
Figure 4.9: experimental setup: steel beam with ends embedded in sand
piezoelectric patches of different thicknesses (1mm, 8mm) were used, all of them
made from PIC141. The material data for PIC141 is given in table 4.2. A stan-
dard cyanoacrylate fast curing adhesive2) was used to bond the PZT patches to
the steel beams. Care must be taken to press the patch firmly against the beam
surface during bonding to attain thin bond gaps. The material properties of the
bonding layer were estimated3) to Gb = 3 ∗ 10−7[ Nm2 ] for the shear stiffness and
εr = 4 for the relative dielectric permittivity.
Admittance curves of the electromechanical system were measured for the
free (supported by foam rubber) and embedded beam, using an impedance
2)LOCTITE4963)Material data for this product is available on the internet at http://www.loctite.com today,
but the estimated values are sufficiently close to the exact values considering that the bondinglayer thickness is unknown.
4.2. Power Factor: Theory versus Experiment for an infinite beam 49
PIC141 material data
quantity variable value unit
mass density ρp 7750 kgm3
quality factor Q 1500 1
loss tangent tan δ 0.005 1
permittivity εT33 1400 ∗ ε0 CVm
compliance sE11 1.26 ∗ 10−11 m2
N
piez. const. d31 −1.27 ∗ 10−10 mV
Table 4.2: material data for PIC141
analyzer4) at 1V output voltage. It turned out, that the longitudinal and flexural
waves were damped out by the sand very well in a frequency range between 500
and 3000 Hz, so the infinite beam could be simulated in this frequency range in
our experimental setup.
Fig. 4.10 shows admittance curves for the free beam. The top dashed curve
is the computed absolute value |Y | of the admittance for the free beam with
zero bonding layer. Measured data is represented by the dotted line, from which
the thickness of the bonding layer was identified, such that the solid line, the
computed |Y | for finite bonding layer thickness, gives as good as possible an
approximation of the measurement. A thickness of s = 2.5µm was identified and
it should be clear that this value, though still of a reasonable order of magnitude,
will generally not be the actual thickness of the layer. However, the bonding layer
thickness is the only parameter used to fit measured and calculated data, and it
is unique to each bonded actuator, for all of the computations carried out. The
resonance peaks in fig. 4.10 are mostly flexural modes. In the frequency range
observed, only two longitudinal resonances occur for the 3m steel beam.
Fig. 4.11 shows the curves corresponding to fig. 4.10, where now the beam
ends are embedded in sand and numerical results are plotted for the infinite beam
model. Again, for the identified 2.5µm thickness of the bonding layer, measured
(dotted line) and computed (solid line) values for |Y | are in very good agreement,
which is not too surprising. Apparently, longitudinal and flexural resonances are
suppressed by the sand bed well enough to achieve a good approximation of the
behavior of an infinite beam.
4)Hewlett Packard 4194A
50 Chapter 4. Verification
500 1000 1500 2000 2500 30000
1
2
3
4
5
x 10−4
f [Hz]
|Y| [
S]
Figure 4.10: absolute value of admittance for free beam: h = 1mm, H = 5mm,
s = 2, 5µm; (– –)zero bonding layer thickness, (—)numerical data, (· · ·)measured data
It is rather the capacitive influence of the bonding layer than the mechanical,
that causes the shift in the simulated admittance curves. Still, the mechanical
aspects of the actuator’s behavior are well represented. This is evident from fig.
4.12, where the computed displacement of the beam in 100mm distance from the
boundary of the laminate section is plotted over the measured values for the same
identified bonding layer thickness of 2.5µm. A two channel laser vibrometer5)
was used to measure the beam displacement. The reflective stickers, which are
required at the measurement points in connection with the glass fibre optics,
can be seen in fig. 4.9.
In the low frequency regime, a strong deviation of measured and computed
data can be observed. A possible explanation would be the near field of the
bending waves originating from the discontinuity at the boundary of the lam-
inate section. For low frequencies, these near fields reach into the sand cov-
ered area which might lead to undesired reflections and resonance phenomena.
This low frequency deviation from the ideal behavior might not be evident in
the admittance curves, because admittance measurements were performed at
5)POLYTEC OFN508/2802
4.2. Power Factor: Theory versus Experiment for an infinite beam 51
0 500 1000 1500 2000 2500 30000
1
2
3
4
5
x 10−4
f [Hz]
|Y| [
S]
Figure 4.11: absolute value of admittance for embedded beam: h = 1mm, H = 5mm,
s = 2, 5µm; (– –)zero bonding layer thickness, (—)numerical data, (· · ·)measured data
1V output voltage whereas a 100V amplitude was required for the displacement
measurement to obtain a reasonable signal to noise ratio for the laser vibrometer
output. Above 500Hz the results are still nice.
Similar good agreement of measurements and computed results is observed
for the power factor cosφ. For an 8mm PZT patch on a 3mm steel beam,
measured (circles) and computed (solid line) data are plotted in fig. 4.13. The
identified bonding layer thickness in this case was s = 2.5µm.
4.2.4 Optimal patch geometry
For the 70mm length actuator patches used, it was assumed they would be ideally
suited to excite flexural waves of 140mm wavelength in the substrate beam. At
the corresponding frequency of excitation, 1431Hz for a 3mm steel beam, the
steady state response of the single patch on an infinite beam was calculated for
various actuator geometries. Only the flexural wave contributions Pout,flex to
the power output were considered for the computation of the effective bending
52 Chapter 4. Verification
0 500 1000 1500 2000 2500 30000
1
2
3
4
5x 10
−6
f [Hz]
|W| [
m]
Figure 4.12: displacement w(x1 = 135mm): theory (—)and experiment (· · ·)
power factor κ according to
κ = ην cosφ =Pout,flex|Pin,elec| , (4.30)
where |Pin,elec| is the apparent electrical input power, η is the efficiency of the
power conversion of the lossy actuator patch and ν is the percentage of the output
mechanical power which is converted into flexural waves. At the considered
frequency of excitation, ν is close to one for an actuator thickness of h=1mm
(fig. 4.8).
An optimum geometry for maximum κ of the actuator patch was determined
for zero bonding layer thickness. In fig. 4.14, the surface κ over length L and
thickness h of the patch is represented by isovalue lines of constant κ. The
optimum can be located at a thickness h∗=8mm and length L∗=170mm of the
patch. It was first surprising, that an actuator of length 170mm should be
ideally suited to excite a bending wave of wavelength 140mm. But considering
the impressive thickness of the optimal sized actuator, the stiffening effect of the
huge PZT patch in the laminate section leads to a wavelength of approximately
twice the actuator length for the given frequency of excitation.
The capacitive and mechanical influence of a changing bonding layer thick-
ness on the optimum geometry L∗, h∗ was investigated. Assuming a conductive
4.2. Power Factor: Theory versus Experiment for an infinite beam 53
0 500 1000 1500 2000 2500 3000−0.01
−0.005
0
0.005
0.01
0.015
0.02
0.025
f [Hz]
cos
φ
Figure 4.13: power factor cosφ: theory (—)and experiment (); h =8mm, H =3mm,
s = 2.5µm
5 10 15 20 25 30
50
100
150
200
250
300
h [mm]
L [
mm
]
0.031158
0.00
4451
2
0.00
4451
2
0.013
354
0.022256
Figure 4.14: isovalues κ=const. over length L and thickness h of the PZT patch;
f=1431Hz, PIC141 on steel beam, s = 0
54 Chapter 4. Verification
conductive bonding layer
s [µm] L∗ [mm] h∗ [mm] κ
0 170 8 0,036
4 170 8 0,035
40 180 11 0,031
400 210 16 0,004
s = 4µm
εr L∗ [mm] h∗ [mm] κ
1 210 13 0,017
4 180 10 0,027
10 175 9 0,033
20 170 8 0,034
Table 4.3: influence of changing layer thickness s and relative permittivity εr on
optimal geometry L∗, h∗ and maximum effective bending power factor κ
bonding layer, the magnitude of the maximum κ and the corresponding geome-
try of the patch is given in table 4.3 for different bonding layer thicknesses. Also,
for a constant bonding layer thickness of s=4µm, the maximum κ and L∗, h∗
are listed for different relative dielectric permittivities εr of the bonding layer.
Presuming that bonding layer thicknesses of s ≤10µm can be achieved, the
mechanical influence of the bonding layer on the optimal design and the result-
ing effective bending power factor is weak. The capacity of the bonding layer
depends not only on the dielectric permittivity of the adhesive, but for a simple
plate capacitor model also is inversely proportional to the thickness of the layer.
Taking into consideration that the manufacturer’s exact value for the relative
permittivity εr of the adhesive is 2.75 and the assumed bonding layer thickness
of s=4µm might be on the optimistically thin edge of the achievable range, it
can be said that the capacitive influence of the bonding layer on the performance
of the actuator, i.e. the maximum effective bending power factor is strong. The
bonding layer should be as thin as possible and the permittivity of the layer and
therefore the resulting capacity Cb should be high. In some applications it might
be advised to use conductive epoxy for bonding, which eliminates any capacitive
influence of the bonding layer. However, the cyanoacrylate adhesive used in the
experiments gave satisfactory results if the actuator patches were firmly pressed
against the substrate beam for the curing time.
Chapter 5
Discussion and Conclusion
5.1 Discussion of the results
Wang et al. [WDXC99] computed an ’electromechanical coupling coefficient’
of a cantilevered piezoelectric bimorph1) actuator for different thickness ratios of
substrate and PZT layer. The origin of the definition of this coupling coefficient
remained somewhat unclear, still the resulting plots agree with figs. 4.3 and 4.2
and optimal thicknesses for the combination of PZT/steel and PZT/aluminum
agree with results obtained from this work, whereas the computed coupling
coefficients in the work of Wang et al. are smaller.
In experiments with cantilever bimorphs, Cunningham et al. [CJB97] de-
termined the resonant amplitude of vibration for the first eigenmode of the
cantilever for varying thicknesses of the substrate layer under constant voltage
amplitude of excitation. The resulting amplitude of vibration at the free end of
the cantilever exhibits the same dependence on the thickness ratio of the two
layers as the EMCF in figs. 4.3 and 4.2. In particular, optimum thickness ratios
for steel and aluminum substrate layers are comparable to the ones obtained
from EMCF considerations in this work.
Strambi and Barboni [SBG95] computed the normalized curvature of a
Bernoulli–Euler laminate, depending on the substrate to actuator thickness
ratio. Their plotted results show close similarity to the corresponding curves for
the EMCF, presented in figs. 4.3 and 4.2. Again, stiffer substrate materials lead
to thicker piezoelectric layers for optimal curvature.
1)A bimorph is considered a laminated beam consisting of any two different layers extendingover the entire length of the beam.
56 Chapter 5. Discussion and Conclusion
5.2 Conclusion
For piezoelectric bending actuators, an attempt was made to apply general crite-
ria to optimize the piezoelectric layer thickness for best energy conversion. This
attempt has been partially successful. It was shown, that the electromechanical
coupling factor (EMCF) can be employed to determine optimal thicknesses and
that these optimal values are independent of the prescribed deformation. Based
on a quasi-static load cycle, the EMCF is a measure for the relative amount
of the energy supplied to the system in one cycle, which is actually converted.
This quasi-static load cycle is of hypothetical character and not a priori repre-
sentative of the energy conversion in systems undergoing steady-state harmonic
vibrations. It is the power factor cosφ, which gives the ratio of the converted
to the maximally supplied energy in one harmonic cycle, and therefore can be
considered the equivalent to the EMCF for steady-state vibration of electrome-
chanical systems. Unlike it was the case for the EMCF, analytical expressions
for an optimal thickness of the piezoelectric layer based on the actuator power
factor could not be given. A model for a single piezoelectric actuator on an
infinite beam was presented. Comparing numerical solutions to measured data
obtained from an experimental setup, it was shown that this is a good model
for the energy conversion in piezoelectric bending actuators. An optimal length
and thickness for maximum power factor of the single actuator patch could be
determined.
It is evident from both, the optimization for maximum EMCF and for max-
imum power factor in the case of the single actuator on the infinite beam, that
the optimal thicknesses for the piezoelectric layer can be comparable to the
thickness of the substrate layer, if the substrate material is stiff. Typical layer
thicknesses in current technical applications are smaller. Of course, the thickness
of the piezoelectric layer may also be limited by the strength of the material, as
thicker layers will generally be expected to exhibit larger strains under similar
deformation. However, if material strength is not a concern, current designs of
piezoelectric layers for bending actuation should be reconsidered.
Bibliography
[Aul81] B.A. Auld: Wave Propagation and Resonance in Piezoelectric Ma-
terials. J. Acoust. Soc. Am, 70(6), 1577–1585, 1981.
[Aul89] B. A. Auld: Acoustic Fields and Waves in Solids. Krieger Pub-
lishing Company, 1989.
[BCJ] Don A. Berlincourt, Daniel R. Curran, Hans Jaffe: Piezo-
electric and Piezomagnetic Materials and Their Function in Trans-
ducers, chapter 3.
[BSW95] H.T. Banks, R.C. Smith, Yun Wang: The Modeling of Piezoce-
ramic Patch Interactions with Shells, Plates and Beams. Quarterly
of Applied Mechanics, 53(2), 353–381, 1995.
[CA90] Edward F. Crawley, Eric H. Anderson: Detailed Models of
Piezoceramic Actuation of Beams. Journal of Intelligent Material
Systems and Structures, 1, 5–25, 1990.
[CdL87] Edward F. Crawley, Javier de Luis: Use of Piezoelectric Actu-
ators as Elements of Intelligent Structures. AIAA Journal, 25(10),
1373–1385, 1987.
[CH88] L. Cremer, M. Heckl: Structure-Borne Sound, New York.
Springer, 1988.
[CJB97] M.J. Cunningham, D.F.L. Jenkins, M.M. Bakush: Experimen-
tal Investigation of Optimum Thickness of a Piezoelectric Element
for Cantilever Actuation. volume 144 671–681, 1997.
[CR94] Z. Chaudry, C.A. Rogers: The Pin-Force Model Revisited. Jour-
nal of Intelligent Material Systems and Structures, 5, 347–354, 1994.
58 Bibliography
[Fre97] Stephan Frese: Electromechanical Coupling in Laminate Piezo-
ceramic Beam Structures. Master’s thesis, Darmstadt University of
Technology, Department of Applied Mechanics, 1997.
[GF92] Gary P. Gibbs, Chris R. Fuller: Excitation of Thin Beams
using Asymmetric Piezoelectric Actuators. J. Acoust. Soc. Am., 6,
3221–3227, 1992.
[GUS89] V. T. Grinchenko, A. F. Ulitko, N. A. Shul’ga: Mechanics
of Coupled Fields in Elements of Structures, Vol 5: Electroelasticity
(in Russian), Kiev. Naukova Dumka, 1989.
[HN68] Richard Holland, E.P. Eer Nisse: Variational Evaluation
of Admittances of Multielectroded Three-Dimensional Piezoelectric
Structures. IEEE Transactions on Sonics and Ultrasonics, SU-15(2),
119–132, 1968.
[HN69] R. Holland, E.P. Eer Nisse: Design of Resonant Piezoelectric
Devices, Cambridge, Massachusetts. The M.I.T. Press, 1969.
[HPS+94] Craig L. Hom, Steven M. Pilgrim, Natarajan Shankar,
Keith Bridger, Mona Massuda, Stephen R. Winzer: Cal-
culation of Quasi-Static Electromechanical Coupling Coefficients for
Electrostrictive Ceramic Materials. IEEE Transactions on Ultrason-
ics, Ferroelectrics and Frequency Control, 41(4), 542–551, 1994.
[Hun54] F.V. Hunt: Electroacoustics, New York. John Wiley & Sons, Inc.,
1954.
[IEE78] ANSI/IEEE Std. 176-1978: IEEE Standard on Piezoelectricity, New
York. The Institute of Electrical and Electronics Engineers, Inc.,
1978.
[IEE88] ANSI/IEEE Std. 176-1987: IEEE Standard on Piezoelectricity, New
York. The Institute of Electrical and Electronics Engineers, Inc.,
1988.
[IRE58] IRE Standards on Piezoelectric Crystals: Determination of the Elas-
tic, Piezoelectric, and Dielectric Constants - The Electromechanical
Coupling Factor, 1958. volume 46 764–778, 1958.
[JCJ71] B. Jaffe, W. R. Cook, H. Jaffe: Piezoelectric Ceramics, Mari-
etta, OH. R.A.N. Publishers, 1971.
Bibliography 59
[JS93] Franco Jona, G. Shirane: Ferroelectric Crystals, New York.
Dover Publications, 1993.
[KJ91] S. J. Kim, J. D. Jones: Optimal Design of Piezoactuators for
Active Noise and Vibration Control. AIAA Journal, 29(12), 2047–
2053, 1991.
[Ler90] Reinhard Lerch: Simulation of Piezoelectric Devices by Two- and
Three-Dimensional Finite Elements. IEEE Transactions on Ultra-
sonics, Ferroelectrics and Frequency Control, 37(2), 233–247, 1990.
[LP91] N. Lamberti, M. Pappalardo: Computation of the Electrome-
chanical Coupling Coefficient k in Piezoelectric Bounded Plates by
Means of a Two-Dimensional Model. Ultrasonics Symposium 1991,
volume 1 621–624, 1991.
[LSR95] C. Liang, F.P. Sun, C.A. Rogers: Determination of Design
of Optimal Actuator Location and Configuration Based on Actuator
Power Factor. Journal of Intelligent Material Systems and Struc-
tures, 6, 456–464, 1995.
[LSR97] C. Liang, F.P. Sun, C.A. Rogers: An Impedance Method for
Dynamic Analysis of Active Material Systems. Journal of Intelligent
Material Systems and Structures, 8, 323–334, 1997.
[Mas50] Warren P. Mason: Piezoelectric Crystals and Their Application
to Ultrasonics. D. Van Nostrand Company, Inc., 1950.
[Mau88] Gerard A. Maugin: Continuum Mechanics of Electromagnetic
Solids, New York, volume 33 of Applied Mathematics and Mechanics.
Elsevier Science Publishing Company, 1988.
[Ngu99] Minh Nam Nguyen: Nichtlineares dynamisches Verhalten von
Piezo-Balken-Systemen bei schwachem elektrischem Feld. PhD the-
sis, Darmstadt University of Technology, 1999.
[Nis67] Errol P. Eer Nisse: Variational Method for Electroelastic Vibra-
tion Analysis. IEEE Transactions on Sonics and Ultrasonics, SU-
14(4), 153–160, 1967.
[Rei96] Howard Reiss: Methods of Thermodynamics, Mineola, N.Y.
11501. Dover Publications, 1996.
[SBG95] Giulio Strambi, Renato Barboni, Paolo Gaudenzi: Pin-
Force and Euler-Bernoulli Models for Analysis of Intelligent Struc-
tures. AIAA Journal, 33(9), 1746–1749, 1995.
60 Bibliography
[Sto98] Boris Stober: Optimierung piezoelektrischer Aktuatoren zur
Erzeugung von Biegeschwingungen. Master’s thesis, Darmstadt Uni-
versity of Technology, Department of Applied Mechanics, 1998.
[Tie68] H.F. Tiersten: Natural Boundary and Initial Conditions from
a Modification of Hamilton’s Principle. Journal of Mathematical
Physics, 9(9), 1445–1451, 1968.
[Tie69] H. F. Tiersten: Linear Piezoelectric Plate Vibrations, New York.
Plenum Press, 1969.
[Uli77] A.F. Ulitko: Theory of Electromechanical Energy Conversion in
Nonuniformly Deformable Piezoceramics. Soviet Applied Mechan-
ics; cover to cover translation of Prikladnaya Mekhanika 1055–1062,
1977.
[UTKN93] S. Ueha, Y. Tomikawa, M. Kurosawa, N. Nakamura: Ultra-
sonic Motors, Theory and Applications, Oxford. Clarendon Press,
1993.
[Voi28] W. Voigt: Lehrbuch der Kristallphsik, Leipzig. Teubner-Verlag,
1928.
[Was68] K. Washizu: Variational Methods in Elasticity and Plasticity, Ox-
ford. International Series of Monographs in Aeronautics and Astro-
nautics. Pergamon Press, 1968.
[WDXC99] Quing-Ming Wang, Xiao-Hong Du, Baomin Xu, L. Eric
Cross: Electromechanical Coupling and Output Efficiency of Piezo-
electric Bending Actuators. IEEE Transactions on Ultrasonics, Fer-
roelectrics and Frequency Control, 46(3), 638–646, 1999.
[WFHS97] K. Wolf, S. Frese, P. Hagedorn, W. Seemann: Thickness-
Optimization of Piezoceramic Transducers for Energy Transfer. Pro-
ceedings of DETC’97, 1997.
[ZR95] S. Zhou, C. A. Rogers: Power Flow and Consumption in Piezo-
electrically Actuated Structures. AIAA Journal, 33, 1305–1311,
1995.
Lebenslauf
Kai-Dietrich Wolf
Karlstraße 30
65185 Wiesbaden
- Geboren am 29.03.68 in Wiesbaden
Mai 1987 Abiturprufung an der Oranienschule in Wiesbaden
Okt 1988 Immatrikulation an der TUD, Fachbereich Maschinenbau
Jun 1991 Abschluß des Vordiploms Maschinenbau
Okt 1991 Wechsel in den Fachbereich Mechanik
Aug 1992 Integriertes Auslandsstudium an der ’University of
California’, Berkeley (1 Jahr)
Sep 1994 Abschlußprufung Mechanik
seit Okt 1994 Wissenschaftlicher Mitarbeiter am Fachbereich Mechanik,
AG Dynamik der Technischen Universitat Darmstadt